d50311gc40 Rac Admin SG

Unauthorized reproduction or distribution prohibited Copyright 2013, Oracle and/or its affiliates
s a
) ha
o m
y s c e
u n is uid
Oracle Database
b
r 11g: n t GRAC
n t o@ tude
Administration
c i me this S
n as use
e r ton e Student
to Guide
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non
D50311GC40
Edition 4.0
December 2012
D78200
Authors Copyright 2012, Oracle and/or its affiliates. All rights reserved.
James Womack Disclaimer
James Spiller This document contains proprietary information and is protected by copyright and
other intellectual property laws. You may copy and print this document solely for your
own use in an Oracle training course. The document may not be modified or altered
Technical Contributors in any way. Except where your use constitutes "fair use" under copyright law, you
and Reviewers may not use, share, download, upload, copy, print, display, perform, reproduce,
publish, license, post, transmit, or distribute this document in whole or in part without
Christopher Andrews the express authorization of Oracle.
Christian Bauwens
The information contained in this document is subject to change without notice. If you
Harald Van Breederode find any problems in the document, please report them in writing to: Oracle University,
David Brower 500 Oracle Parkway, Redwood Shores, California 94065 USA. This document is not
warranted to be error-free.
Michael Cebulla
Jonathan Creighton Restricted Rights Notice
Al Flournoy If this documentation is delivered to the United States Government or anyone using
Andy Fortunak the documentation on behalf of the United States Government, the following notice is
Mark Fuller
applicable:
s a
Joel Goodman U.S. GOVERNMENT RIGHTS ) ha
o m
The U.S. Governments rights to use, modify, reproduce, release, perform, display, or
Michael Hazel
s c e
disclose these training materials are restricted by the terms of the applicable Oracle
y
Pete Jones
is uid
license agreement and/or the applicable U.S. Government contract.
u n
Mike Leatherman
Trademark Notice
b r n t G
o@ tude
Jerry Lee
Barb Lundhild n t
Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names
Markus Michalewicz
c i me this S
may be trademarks of their respective owners.
Peter Sharman
n as use
Ranbir Singh
e r ton e to
Linda Smalley
( e v ens
Janet Stern
o n n e lic
r t
Richard Strohm
e a b l
Ev nsfer
S. Matt Taylor
n
e r to -tra
Branislav Valny
Ev non
Jean-Francois Verrier
Rick Wessman
Doug Williams
Editors
Aju Kumar
Smita Kommini
Daniel Milne
Publishers
Jobi Varghese
Sujatha Nagendra
Contents
1 Grid Infrastructure: Overview

Objectives 1-2
Oracle Grid Infrastructure 1-3
What Is a Cluster? 1-4
What Is Clusterware? 1-5
Oracle Clusterware 1-6
Oracle Clusterware Architecture and Services 1-7
s a
Goals for Oracle Clusterware 1-8 ) ha
Oracle Clusterware Networking 1-9 o m
Oracle Grid Infrastructure for a Cluster 1-11 y s c e
u n is uid
Oracle Clusterware Initialization 1-12
b r n t G
Clusterware Startup Details 1-13
n t o@ tude
me this S
Clusterware Startup: The OHASD orarootagent 1-15
c i
Clusterware Startup Details: The CRSD orarootagent 1-17
n as use
Clusterware Startup Details: The CRSD oraagent 1-18
r ton e to
Clusterware Startup Details: The OHASD oraagent 1-19
e
( e v ens
Grid Plug and Play 1-20
o n n e lic
e r t a b l
Grid Naming Service 1-21
n Ev nsfer
Single Client Access Name 1-22
e r to -tra
Oracle Automatic Storage Management (ASM) 1-24
Ev non ASM Key Features and Benefits 1-25
ASM and Grid Infrastructure 1-26
Quiz 1-27
Summary 1-30
Practice 1 Overview 1-31
2 RAC Concepts
Objectives 2-2
Overview of Oracle RAC 2-3
RAC One Node Single-Instance High Availability 2-5
Oracle RAC One Node 2-6
Oracle RAC One Node and Oracle Clusterware 2-7
Cluster-Aware Storage Solutions 2-8
Oracle Cluster File System 2-9
Benefits of Using RAC 2-10
iii
Clusters and Scalability 2-11
Levels of Scalability 2-12
Scaleup and Speedup 2-14
Speedup/Scaleup and Workloads 2-15
I/O Throughput Balanced: Example 2-16
Performance of Typical Components 2-17

Necessity of Global Resources 2-18
Additional Memory Requirement for RAC 2-19
Parallel Execution with RAC 2-20
Summary 2-21
3 Installing and Configuring Oracle RAC a

Objectives 3-2 ha s
Installing the Oracle Database Software 3-3 m )
o
c e
Creating the Cluster Database 3-8 y s
Database Type Selection 3-9 u n is uid
Database Identification 3-10 b r n t G
n
Cluster Database Management Options 3-11 t o@ tude
i me this S
Passwords for Database Schema Owners 3-12
c
Database File Locations 3-13
n as use
e r ton e to
Recovery Configuration 3-14
e v ens
Database Content 3-15
(
n n e lic
Initialization Parameters 3-16
o
e r t a b l
n Ev nsfer
Database Storage Options 3-17
Create the Database 3-18
e r to -tra
Ev non Monitoring Progress 3-19
Postinstallation Tasks 3-20
Checking Managed Targets 3-21
Background Processes Specific to Oracle RAC 3-22
Single Instanceto-RAC Conversion 3-24
Considerations for Converting Single-Instance Databases to Oracle RAC 3-25
Single-Instance Conversion Using the DBCA 3-26
Conversion Steps 3-27
Single-Instance Conversion Using rconfig 3-30
Quiz 3-32
Summary 3-34
4 Oracle RAC Administration

Objectives 4-2
Cluster Database Home Page 4-3
iv
Cluster Database Instance Home Page 4-5
Cluster Home Page 4-6
Configuration Section 4-7
Topology Viewer 4-9
Enterprise Manager Alerts and RAC 4-10
Enterprise Manager Metrics and RAC 4-11

Enterprise Manager Alert History and RAC 4-13
Enterprise Manager Blackouts and RAC 4-14
Redo Log Files and RAC 4-15
Automatic Undo Management and RAC 4-16
Starting and Stopping RAC Instances 4-17
Starting and Stopping RAC Instances with srvctl 4-18 a
Starting and Stopping RAC Instances with SQL*Plus 4-19 ha s
Switch Between Automatic and Manual Policies 4-20 m )
o
c e
RAC Initialization Parameter Files 4-21 y s
SPFILE Parameter Values and RAC 4-22 u n is uid
EM and SPFILE Parameter Values 4-23 b r n t G
RAC Initialization Parameters 4-25 n t o@ tude
i me this S
Parameters That Require Identical Settings 4-27
c
n as use
Parameters That Require Unique Settings 4-28

e r ton e to
Quiescing RAC Databases 4-29
e v ens
Terminating Sessions on a Specific Instance 4-30
(
n n e lic
How SQL*Plus Commands Affect Instances 4-31
o
e r t a b l
n Ev nsfer
Transparent Data Encryption and Wallets in RAC 4-32
Quiz 4-33
e r to -tra
Ev non Summary 4-35
5 Managing Backup and Recovery for RAC

Objectives 5-2
RAC and Instance Recovery 5-3
Instance Recovery and Database Availability 5-5
Instance Recovery and RAC 5-6
Protecting Against Media Failure 5-8
Media Recovery in Oracle RAC 5-9
Parallel Recovery in RAC 5-10
Archived Log File Configurations 5-11
RAC and the Fast Recovery Area 5-12
RAC Backup and Recovery Using EM 5-13
Configuring RAC Recovery Settings with EM 5-14
Archived Redo File Conventions in RAC 5-15
v
Configuring RAC Backup Settings with EM 5-16
Oracle Recovery Manager 5-17
Configuring RMAN Snapshot Control File Location 5-18
Configuring Control File and SPFILE Autobackup 5-19
Crosschecking on Multiple RAC Clusters Nodes 5-20
Channel Connections to Cluster Instances 5-21

RMAN Channel Support for the Grid 5-22
RMAN Default Autolocation 5-23
Distribution of Backups 5-24
Shared Storage Backup Scheme: One Local Drive 5-25
Shared Storage Backup Scheme: Multiple Drives 5-26
Restoring and Recovering 5-27 a
Quiz 5-28 ha s
Summary 5-30 m )
o
c e
Practice 5 Overview 5-31 y s
Objectives 6-2 u n is uid
Need for Global Concurrency Control 6-3 b r n t G
Global Resource Directory (GRD) 6-4 n t o@ tude
Global Resource Management 6-5 c i me this S
n
Global Resource Remastering 6-6
as use
e r ton e to
Global Resource Recovery 6-7
e v ens
Global Resource Background Processes 6-8
(
n n e lic
Global Resource Access Coordination 6-10
o
e r t a b l
n Ev nsfer
Global Enqueues 6-11
Instance Locks 6-12
e r to -tra
Ev non Global Cache Management: Overview 6-13
Global Cache Management Components 6-14
Global Cache Buffer States 6-15
Global Cache Management Scenarios for Single Block Reads 6-16
Global Cache Scenarios: Overview 6-17
Scenario 1: Read From Disk 6-18
Scenario 2: Read-Write Cache Fusion 6-22
Scenario 3: Write-Write Cache Fusion 6-26
Scenario 4: Write-Read Cache Fusion 6-30
Global Cache Management Scenarios for Multi-Block Reads 6-34
Useful Global Resource Management Views 6-35
Quiz 6-36
Summary 6-37
vi
7 RAC Database Monitoring and Tuning
Objectives 7-2
CPU and Wait Time Tuning Dimensions 7-3
RAC-Specific Tuning 7-4
Analyzing Cache Fusion Impact in RAC 7-5
Typical Latencies for RAC Operations 7-6

Wait Events for RAC 7-7
Wait Event Views 7-8
Global Cache Wait Events: Overview 7-9
Global Enqueue Waits 7-11
Session and System Statistics 7-12
Most Common RAC Tuning Tips 7-13 a
Index Block Contention: Considerations 7-15 ha s
Oracle Sequences and Index Contention 7-16 m )
o
c e
Undo Block Considerations 7-17 y s
High-Water Mark Considerations 7-18 u n is uid
b r
Concurrent Cross-Instance Calls: Considerations 7-19 n t G
n t o@ tude
Monitoring RAC Database and Cluster Performance 7-20
i me this S
Cluster Database Performance Page 7-21
c
n as use
Determining Cluster Host Load Average 7-22

e r ton e to
Determining Global Cache Block Access Latency 7-23
e v ens
Determining Average Active Sessions 7-24
(
n n e lic
Determining Database Throughput 7-25
o
e r t a b l
n Ev nsfer
Accessing the Cluster Cache Coherency Page 7-27
Viewing the Cluster Interconnects Page 7-29
e r to -tra
Ev non Viewing the Database Locks Page 7-31
AWR Snapshots in RAC 7-32
AWR Reports and RAC: Overview 7-33
Active Session History Reports for RAC 7-35
Automatic Database Diagnostic Monitor for RAC 7-37
What Does ADDM Diagnose for RAC? 7-39
EM Support for ADDM for RAC 7-40
Quiz 7-41
Summary 7-43
8 Managing High Availability of Services

Objectives 8-2
Oracle Services 8-3
Services for Policy- and Administrator-Managed Databases 8-4
Default Service Connections 8-5
vii
Creating Service with Enterprise Manager 8-6
Creating Services with SRVCTL 8-7
Managing Services with Enterprise Manager 8-8
Managing Services with EM 8-9
Managing Services with srvctl 8-10
Using Services with Client Applications 8-11

Services and Connection Load Balancing 8-12
Services and Transparent Application Failover 8-13
Using Services with the Resource Manager 8-14
Services and Resource Manager with EM 8-15
Using Services with the Scheduler 8-16
Services and the Scheduler with EM 8-17 a
Using Distributed Transactions with RAC 8-19 ha s
Distributed Transactions and Services 8-20 m )
o
c e
Service Thresholds and Alerts 8-22 y s
Services and Thresholds Alerts: Example 8-23 u n is uid
Service Aggregation and Tracing 8-24 b r n t G
Top Services Performance Page 8-25 n t o@ tude
Service Aggregation Configuration 8-26 c i me this S
n as use
Service, Module, and Action Monitoring 8-27

Service Performance Views 8-28
e r ton e to
Quiz 8-29 ( e v ens
Summary 8-31 o n n e lic
e r t a b l
Ev nsfer
n
e r toHigh-Availability
tra
E v 9
o n of Connections
nObjectives 9-2
Types of Workload Distribution 9-3
Client-Side Connect-Time Load Balancing 9-4
Client-Side Connect-Time Failover 9-5
Server-Side Connect-Time Load Balancing 9-6
Fast Application Notification: Overview 9-7
Fast Application Notification: Benefits 9-8
FAN-Supported Event Types 9-9
FAN Event Status 9-10
FAN Event Reasons 9-11
FAN Event Format 9-12
Load Balancing Advisory: FAN Event 9-13
Server-Side Callouts Implementation 9-14
Server-Side Callout Parse: Example 9-15
Server-Side Callout Filter: Example 9-16
viii
Server-Side ONS 9-17
Optionally Configuring the Client-Side ONS 9-18
UCP JDBC Fast Connection Failover: Overview 9-19
Using Oracle Streams Advanced Queuing for FAN 9-20
JDBC/ODP.NET FCF Benefits 9-21
Load Balancing Advisory 9-22

UCP JDBC/ODP.NET Runtime Connection Load Balancing: Overview 9-23
Connection Load Balancing in RAC 9-24
Load Balancing Advisory: Summary 9-25
Monitoring LBA FAN Events 9-26
FAN Release Map 9-27
Transparent Application Failover: Overview 9-28 a
TAF Basic Configuration Without FAN: Example 9-29 ha s
TAF Basic Configuration with FAN: Example 9-30 m )
o
c e
TAF Preconnect Configuration: Example 9-31 y s
TAF Verification 9-32 u n is uid
FAN Connection Pools and TAF Considerations 9-33 b r n t G
Summary 9-34 n t o@ tude
c i me this S
10 Upgrading and Patching Oraclen as use
RAC
Objectives 10-2
e r ton e to
Types of Patches 10-3( e v ens
Patch Properties
o n n 10-5 e lic
e r a b l Library 10-6
tthe Software
n EvUpnPatching
Configuring
Setting s fer 10-7
e r toObtaining
- traOracle RAC Patches 10-8
E v o n
nDownloading Patches 10-11
Reduced Down-Time Patching for Cluster Environments 10-12
Rolling Patches 10-13
Out-of-Place Database Upgrades 10-14
Out-of-Place Database Upgrade with OUI 10-15
OPatch: General Usage 10-16
Before Patching with OPatch 10-17
OPatch Automation 10-18
OPatch Automation Examples 10-19
Quiz 10-21
Summary 10-22
Lesson 10 Practice Overview 10-23
ix
11 Oracle RAC One Node
Objectives 11-2
Verifying an Existing RAC One Node Database 11-3
Oracle RAC One Node Online Migration 11-4
Online Migration Considerations 11-5
Performing an Online Migration 11-6

Online Migration Illustration 11-7
Online Maintenance: Rolling Patches 11-10
Adding an Oracle RAC One Node Database to an Existing Cluster 11-12
Converting a RAC One Node Database to RAC 11-13
Converting a Single Instance Database to RAC One Node 11-15
Converting a RAC Database to RAC One Node 11-16 a
Quiz 11-17 ha s
Summary 11-19 m )
o
c e
Lesson 11 Practice Overview 11-20 y s
u n is uid
12 Quality of Service Management b r n t G
Lesson Objectives 12-2 n t o@ tude
QoS Management Background 12-3 cim
e is S
a s e th
QoS Management Overview 12-4
o n n o us
QoS Management and Exadata
e r t t Machine 12-5
Database
e
v
(e l12-6
QoS Management Focus en s
QoS Management n
n Benefits i c
r t o b l e 12-7
ve feraFunctional Overview 12-9
QoS Management
E
t o n Management
QoS
a n s Policy Sets 12-11
r r
n-t Pools 12-12
Eve nServer
o
Performance Classes 12-14
Classification and Tagging 12-16
Performance Policies 12-17
Performance Class Ranks 12-18
Performance Objectives 12-19
Performance Satisfaction Metrics 12-20
Server Pool Directive Overrides 12-21
Overview of Metrics 12-22
QoS Management Architecture 12-24
QoS Management Recommendations 12-25
Implementing Recommendations 12-27
Quiz 12-29
Summary 12-31
Lesson 12 Demonstrations 12-32
x
13 Design for High Availability
Objectives 13-2
Causes of Unplanned Down Time 13-3
Causes of Planned Down Time 13-4
Oracles Solution to Down Time 13-5

RAC and Data Guard Complementarity 13-6
Maximum Availability Architecture 13-7
RAC and Data Guard Topologies 13-8
RAC and Data Guard Architecture 13-9
Data Guard Broker (DGB) and Oracle Clusterware (OC) Integration 13-11
Fast-Start Failover: Overview 13-12
Data Guard Broker Configuration Files 13-14 a
Real-Time Query Physical Standby Database 13-15 ha s
Hardware Assisted Resilient Data 13-16 m )
o
c e
Database High Availability: Best Practices 13-17 y s
How Many ASM Disk Groups Per Database? 13-18 u n is uid
b r
Which RAID Configuration for High Availability? 13-19 n t G
n t o@ tude
Should You Use ASM Mirroring Protection? 13-20
i me this S
What Type of Striping Works Best? 13-21
c
ASM Striping Only 13-22 n as use
e r ton e to
Hardware RAIDStriped LUNs 13-23
e v ens
Hardware RAIDStriped LUNs HA 13-24
(
n n e lic
Disk I/O Design Summary 13-25
o
e r t a b l
n Ev nsfer
Extended RAC: Overview 13-26
Extended RAC Connectivity 13-27
e r to -tra
Ev non Extended RAC Disk Mirroring 13-28
Achieving Quorum with Extended RAC 13-29
Additional Data Guard Benefits 13-30
Using a Test Environment 13-31
Quiz 13-32
Summary 13-33
xi
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non
Grid Infrastructure: Overview
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
l licOracle and/or its affiliates. All rights reserved.
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives
After completing this lesson, you should be able to:

Explain the principles and purposes of clusters

Describe the Oracle Clusterware architecture
Describe how Grid Plug and Play affects Clusterware
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Oracle Database 11g: RAC Administration 1 - 2

Oracle Grid Infrastructure
ASM and Oracle Clusterware are installed into a single home
directory called the Grid Infrastructure home.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EDatabase
n s fe11g Release 2, Automatic Storage Management (ASM) and Oracle
e o
With Oracle
rt n-are trainstalled into a single home directory, collectively called Oracle Grid
v
Clusterware
no This directory is referred to as the Grid Infrastructure home. Configuration
EInfrastructure.
assistants start after the Oracle Universal Installer interview process and binary installation
that configure ASM and Oracle Clusterware. Although the installation is called Oracle Grid
Infrastructure, Oracle Clusterware and Automatic Storage Manager remain separate
components.

What Is a Cluster?
A group of independent, but interconnected, computers

that act as a single system

Usually deployed to
Network
increase availability and Interconnect
performance or Users
to balance a dynamically
changing workload a
a s
m )h
s co
u n u ideNetwork
isy Storage
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
A clustern E ofsafegroup of independent but interconnected computers whose combined
consistsn
e o
rt can trbea applied to a processing task. A common cluster feature is that it should
resources
v n -
Eappearntooan application as though it were a single server. Most cluster architectures use a
dedicated network (cluster interconnect) for communication and coordination between cluster
nodes.
A common cluster architecture for data-intensive transactions and computations is built
around shared disk storage. Shared-nothing clusters use an alternative architecture where
storage is not shared and data must be either replicated or segmented across the cluster.
Shared-nothing clusters are commonly used for workloads that can be easily and predictably
divided into small units that can be spread across the cluster in parallel. Shared disk clusters
can perform these tasks but also offer increased flexibility for varying workloads. Load
balancing clusters allow a single application to balance its workload across the cluster.
Alternatively, in a failover cluster, some nodes can be designated as the primary host for an
application, whereas others act as the primary host for different applications. In a failover
cluster, the failure of a node requires that the applications it supports be moved to a surviving
node. Load balancing clusters can provide failover capabilities but they can also run a single
application across multiple nodes, providing greater flexibility for different workload
requirements. Oracle supports a shared disk cluster architecture providing load balancing and
failover capabilities. In an Oracle cluster, all nodes must share the same processor
architecture and run the same operating system.

What Is Clusterware?
Software that provides various interfaces and services for a

cluster. Typically, this includes capabilities that:

Allow the cluster to be managed as a whole
Protect the integrity of the cluster
Maintain a registry of resources across the cluster
Deal with changes to the cluster
Provide a common view of resources s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eis a term s e
fused
r t o
Clusterware
t r a n to describe software that provides interfaces and services that
ve and
Eenable o -
nsupport a cluster.
n
Different cluster architectures require clusterware that delivers different services. For
example, in a simple failover cluster, the clusterware may monitor the availability of
applications and perform a failover operation if a cluster node becomes unavailable. In a load
balancing cluster, different services are required to support workload concurrency and
coordination.
Typically, clusterware includes capabilities that:
Allow the cluster to be managed as a single entity (not including OS requirements), if
desired
Protect the integrity of the cluster so that data is protected and the cluster continues to
function even if communication with a cluster node is severed
Maintain a registry of resources so that their location is known across the cluster and so
that dependencies between resources is maintained
Deal with changes to the cluster such as node additions, removals, or failures
Provide a common view of resources such as network addresses and files in a file
system

Oracle Clusterware
Oracle Clusterware is:

A key part of Oracle Grid Infrastructure

Integrated with Oracle
Automatic Storage
Management (ASM)
The basis for ASM Cluster
File System (ACFS) s a
) ha
A foundation for Oracle m
Real Application Clusters
o
c e
y s
(RAC) u n is uid
b r n t G
A generalized cluster
n t o@ tude
infrastructure for all kinds ime his S
a s c et
of applications n us
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsisfaekey part of Oracle Grid Infrastructure, which also includes Automatic
Oracle
e o
Clusterware
tra (ASM) and the ASM Cluster File System (ACFS).
rt Management
v
Storage n -
no 11.2, Oracle Clusterware can use ASM for all the shared files required by the
EIn Release
cluster. Oracle Clusterware is also an enabler for the ASM Cluster File System, a generalized
cluster file system that can be used for most file-based data such as documents,
spreadsheets, and reports.
The combination of Oracle Clusterware, ASM, and ACFS provides administrators with a
unified cluster solution that is not only the foundation for the Oracle Real Application Clusters
(RAC) database, but can also be applied to all kinds of other applications.
Note: Grid Infrastructure is the collective term that encompasses Oracle Clusterware, ASM,
and ACFS. These components are so tightly integrated that they are often collectively referred
to as Oracle Grid Infrastructure.

Oracle Clusterware Architecture and Services
Shared disk cluster architecture supporting application

load balancing and failover

Services include:
Cluster management
Node monitoring
Event services
Time synchronization s a
)ha
Network management m
High availability s co
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsprovidesfe a complete set of cluster services to support the shared disk,
Oracle
e t o
Clusterware
rbalancing a architecture of the Oracle Real Application Cluster (RAC) database.
trcluster
v
load
o n -
EOraclenClusterware can also be used to provide failover clustering services for single-instance
Oracle databases and other applications.
The services provided by Oracle Clusterware include:
Cluster management, which allows cluster services and application resources to be
monitored and managed from any node in the cluster
Node monitoring, which provides real-time information regarding which nodes are
currently available and the resources they support. Cluster integrity is also protected by
evicting or fencing unresponsive nodes.
Event services, which publishes cluster events so that applications are aware of
changes in the cluster
Time synchronization, which synchronizes the time on all nodes of the cluster
Network management, which provisions and manages Virtual IP (VIP) addresses that
are associated with cluster nodes or application resources to provide a consistent
network identity regardless of which nodes are available. In addition, Grid Naming
Service (GNS) manages network naming within the cluster.
High availability, which services, monitors, and restarts all other resources as required

Goals for Oracle Clusterware
Easy installation
Easy management
Continuing tight integration with Oracle RAC
ASM enhancements with
benefits for all applications
No additional clusterware
required s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nshas febecome the required clusterware for Oracle Real Application Clusters
Oracle
e oClusterware
rt Oracle a
trDatabase
v
(RAC). n - 11g Release 2 builds on the tight integration between Oracle
no and RAC by extending the integration with Automatic Storage Management
EClusterware
(ASM). The result is that now all the shared data in your cluster can be managed using ASM.
This includes the shared data required to run Oracle Clusterware, Oracle RAC, and any other
applications you choose to deploy in your cluster.
In most cases, this capability removes the need to deploy additional clusterware from other
sources, which also removes the potential for integration issues caused by running multiple
clusterware software stacks. It also improves the overall manageability of the cluster.
Although most of the enhancements to ASM are the subject of later lessons, the next part of
this lesson examines a series of additional Oracle Clusterware capabilities and the benefits
they provide.

Oracle Clusterware Networking
Each node must have at least two network adapters.
Each public network adapter must support TCP/IP.
The interconnect adapter must support:

User Datagram Protocol (UDP) or Reliable Data Socket (RDS)
for UNIX and Linux for database communication
UDP for Windows platforms for database communication
All platforms use Grid Interprocess Communication (GIPc).
Public network
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use Interconnect: Private network
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
o n
Each tnode
E n
mustahavesfeat least two network adapters: one for the public network interface and
v
theeother tr private network interface or interconnect. In addition, the interface names
r forn-the
no with the network adapters for each network must be the same on all nodes. For
Eassociated
example, in a two-node cluster, you cannot configure network adapters on node1 with eth0 as
the public interface, but on node2 have eth1 as the public interface. Public interface names
must be the same, so you must configure eth0 as public on both nodes. You should configure
the private interfaces on the same network adapters as well. If eth1 is the private interface for
node1, then eth1 should be the private interface for node2.
Before starting the installation, on each node, you must have at least two interfaces to
configure for the public and private IP addresses. You can configure IP addresses with one of
the following options:
Oracle Grid Naming Service (GNS) using one static address defined during installation,
which dynamically allocates VIP addresses by using Dynamic Host Configuration
Protocol (DHCP), which must be running on the network. You must select the Advanced
Oracle Clusterware installation option to use GNS.
Static addresses that network administrators assign on a network domain name server
(DNS) or each node. To use the Typical Oracle Clusterware installation option, you must
use static addresses.

For the public network, each network adapter must support TCP/IP.
For the private network, the interconnect must support UDP or RDS for communications to
the database. Grid Interprocess Communication (GIPc) is used for Grid (Clusterware)
interprocess communication. GIPC is a new common communications infrastructure to
replace CLSC/NS. It provides a full control of the communications stack from the operating
system up to whatever client library uses it. The dependency on network services (NS) prior
to 11.2 is removed, but there is still backwards compatibility with existing CLSC clients
(primarily from 11.1). GIPC can support multiple communications types: CLSC, TCP, UDP,
IPC and of course the communication type GIPC.
Use high-speed network adapters for the interconnects and switches that support TCP/IP.
Gigabit Ethernet or an equivalent is recommended.
Each node in a cluster requires a supported interconnect protocol to support Cache Fusion
and TCP/IP to support Clusterware polling. Token Ring is not supported for cluster
interconnects on IBM AIX. Your interconnect protocol must be certified by Oracle for your
platform. s a
) ha
Note: Cross-over cables are not supported for use with Oracle Clusterware interconnects.
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Oracle Grid Infrastructure for a Cluster
Oracle Clusterware
Platform-independent facility for starting, stopping, and

managing clusterwide resources
Oracle Clusterware is composed of two physical stacks:
Cluster Ready Services stack
Oracle High Availability Services stack
Oracle Automatic Storage Management (ASM) a
Platform-independent shared storage solution providing ) ha
s
storage for the Clusterware, databases, and most other c om
application or system requirements i s ys ide
r u n Gu
@ b ent
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe
EGrid Infrastructure
r t o
With Oracle
t r a 11g Release 2, Oracle Automatic Storage Management
ve ASM)
E(Oracle - and Oracle Clusterware are installed into a single home directory, which is
referred ntoonas the Grid Infrastructure home.
Oracle Clusterware consists of two separate stacks: an upper stack anchored by the Cluster
Ready Services (CRS) daemon (crsd) and a lower stack anchored by the Oracle High
Availability Services daemon (ohasd). These two stacks have several processes that
facilitate cluster operations.
The Cluster Ready Services stack manages cluster resources based on the configuration
information that is stored in OCR for each resource. This includes start, stop, monitor, and
failover operations.
The Oracle High Availability Services stack is responsible for monitoring and maintaining high
availability of Oracle ASM and Oracle Clusterware itself.
The installation of the combined products is called Oracle Grid Infrastructure. However,
Oracle Clusterware and Oracle Automatic Storage Management remain separate products.

Oracle Clusterware Initialization
Oracle Clusterware is started by the OS init daemon.
Operating system
init daemon
Oracle Clusterware processes
/sbin/init ohasd.bin cssdmonitor
Clusterware
octssd.bin crsd.bin
startup script
oraagent.bin gipcd.bin
diskmon.bin mdnsd.bin
/etc/init.d/init.ohasd
ocssd.bin gpnpd.bin
evmd.bin evmlogger.bin
s a
cssdagent oraagent.bin
) ha
ologgerd.bin osysmond.bin
o m
ons
s c e
orarootagent.bin
y
Oracle Clusterware installation modifies r/etc/inittab u nis Guid to
restart ohasd in the event of a crash. @ b ent
e n to Stud
# cat /etc/inittab
s c im this
..
n a urun s e>/dev/null 2>&1 </dev/null
ton e to
h1:35:respawn:/etc/init.d/init.ohasd
e r
v ens
( e
o n n e2012,
Copyright
v e rt rab
Duringo n
the
E nsfeof Oracle Clusterware, the init.ohasd startup script is copied to
installation
v e rt n-traThe wrapper script is responsible for setting up environment variables and
/etc/init.d.
no the Oracle Clusterware daemons and processes.
Ethen starting
The Oracle High Availability Services daemon (ohasd) is responsible for starting in proper
order, monitoring, and restarting other local Oracle daemons including the crsd daemon,
which manages clusterwide resources. When init starts ohasd on Clusterware startup,
ohasd starts orarootagent, cssdagent, and oraagent. Some of the high availability
daemons will be running under the root user with real-time priority, and others will be
running under the Clusterware owner with user-mode priorities after they are started. When a
command is used to stop Oracle Clusterware, the daemons will be stopped, but the ohasd
process will remain running.

Clusterware Startup Details
init
ohasd
cssdagent orarootagent oraagent cssdmonitor s a

) ha
o m
y s c e
u n is uid
cssd
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
The Oracle n fe Services daemon (ohasd) is responsible for starting in proper
EHigh Availability
s
e o tra and restarting other local Oracle Clusterware daemons, up through the
rtmonitoring,
v
order, n -
no which in turn manages clusterwide resources.
Ecrsd daemon,
When a cluster node boots, or Clusterware is started on a running clusterware node, the
init process starts ohasd. The ohasd process then initiates the startup of the processes in
the lower, or Oracle High Availability (OHASD) stack.
The cssdagent process is started, which in turn starts cssd. The cssd process
discovers the voting disk either in ASM or on shared storage and then joins the cluster.
The cssdagent process monitors the cluster and provides I/O fencing. This service
formerly was provided by Oracle Process Monitor Daemon (oprocd). A cssdagent
failure may result in Oracle Clusterware restarting the node.
The orarootagent is started. This process is a specialized oraagent process that
helps crsd start and manage resources owned by root, such as the network and the
grid virtual IP address.

The oraagent process is started. It is responsible for starting processes that do not
need to be run as root.
The oraagent process extends clusterware to support Oracle-specific requirements
and complex resources. This process runs server callout scripts when FAN events
occur. This process was known as RACG in Oracle Clusterware 11g Release 1 (11.1).
The cssdmonitor is started and is responsible for monitoring the cssd daemon.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Clusterware Startup: The OHASD orarootagent
ohasd
cssdagent orarootagent oraagent cssdmonitor
s a
) ha
o m
y
ACFS s c e
osysmond ologgerd crsd diskmon
u n is uictssdd

br ent
DriversG
@
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
The OHASD n n fe
Eorarootagent
s
e o
rt n-tra
v
The process is responsible for starting the following processes:

no
E orarootagent
osysmond: The system monitor service (osysmond) is the monitoring and operating
system metric collection service that sends the data to the cluster logger service,
ologgerd. The cluster logger service receives the information from all the nodes and
persists in the Cluster Health Monitor (CHM) repository. There is one system monitor
service on every node.
ologgerd: There is one cluster logger service (ologgerd) on only one node in a
cluster and another node is chosen by the cluster logger service to house the standby
for the master cluster logger service. If the master cluster logger service fails, the node
where the standby resides takes over as master and selects a new node for standby.
The master manages the operating system metric database in the CHM repository and
interacts with the standby to manage a replica of the master operating system metrics
database.
crsd: The Cluster Ready Services (CRS) process is the primary program for managing
high availability operations in a cluster. The CRS daemon (crsd) manages cluster
resources based on configuration information stored in OCR for each resource. This
includes start, stop, monitor, and failover operations. The crsd process generates
events when the status of a resource changes. When Oracle RAC is installed, the crsd
process monitors the Oracle database components and automatically restarts them
when a failure occurs.

diskmon: The diskmon process monitors and performs I/O fencing for Oracle Exadata.
ACFS Drivers: These drivers are loaded in support of ASM Dynamic Volume Manager
(ADVM) and ASM Cluster File System (ACFS).
ctssd: The Cluster Time Synchronization Service process provides time
synchronization for the cluster in the absence of ntpd. If ntpd is configured, ctssd will
run in observer mode.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Clusterware Startup Details:
The CRSD orarootagent
ohasd
ACFS
Drivers ctssd s a
)ha
co m
s
orarootagent oraagent unis
y
u ide
b r nt G
n t o@ tude
c i me this S
Node vip SCAN n as usnetwork
vip e GNS vip
n
rto se t o
v e n and/or its affiliates. All rights reserved.
n (e 2012,
Copyright l i c eOracle
r t o n le
e a b
The CRSD n Ev nsfer
e r to -tra
orarootagent
v crsdoprocess
EThe n starts another orarootagent process and another oraagent process.
n
The new orarootagent process is responsible for starting the following resources:
Node vip: The node vip is a node application (nodeapp) responsible for eliminating
response delays (TCP timeouts) to client programs requesting a connection to the
database. Each node vip is assigned an unused IP address. This is usually done via
DHCP but can be manually assigned. There is initially one node vip per cluster node at
Clusterware startup. When a cluster node becomes unreachable, the node vip is failed
over to a surviving node and redirects connection requests made to the unreachable
node to a surviving node.
SCAN vip: SCAN vips or Single Client Access Name vips are part of a connection
framework that eliminates dependencies on static cluster node names. This framework
allows nodes to be added to or removed from the cluster without affecting the ability of
clients to connect to the database. If GNS is used in the cluster, three SCAN vips are
started on the member nodes by using IP addresses assigned by the DHCP server. If
GNS is not used, SCAN vip addresses for the cluster can be defined in the DNS server
used by the cluster nodes.
Network: Network resources required by the cluster are started.
GNS vip: If GNS is used to resolve client requests for the cluster, a single GNS vip for
the cluster is started. The IP address is assigned in the GNS server used by the cluster
nodes.
The CRSD oraagent
ohasd
ACFS
Drivers ctssd s a
)ha
co m
s
orarootagent oraagent unis
y
u ide
b r nt G
n t o@ tude
i m e is S
ASM SCAN
a s c e thNode Database
ONS
Instances onListener n s
u Listener Instances
r t t o
( e ve ense
o n n e2012,
Copyright
v e rt rab
The CRSD n E nsfe
e o
rt n-tra
oraagent
v
As noted
E noin the previous slide, the crsd process starts another orarootagent process and
another oraagent process. The new oraagent process is responsible for starting the
following resources:
ONS: The ONS or Oracle Notification Service is a publishing and subscribing service for
communicating Fast Application Notification (FAN) events.
ASM Instances: The Oracle ASM instances provide disk management for Oracle
Clusterware and Oracle Database. One ASM instance is started on each cluster node.
SCAN Listener: Three SCAN listeners are started on the cluster nodes where the
SCAN vips are started. Oracle Database 11g Release 2 and later instances only register
with SCAN listeners as remote listeners.
Node Listener: If GNS is used to resolve client requests for the cluster, a single GNS
vip for the cluster is started. The IP address is assigned in the GNS server used by the
cluster nodes.
Database Instances: If the cluster nodes are supporting an Oracle RAC database, the
database instances are started.

The OHASD oraagent
ohasd
orarootagent cssdagent oraagent cssdmonitor
s a
)ha
co m
s
gipcd mdnsd evmd ASM nis
u
y gpnpd
u ide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
The OHASD n Eoraagent
n s fe
e o
rt n-tra
v
The

no process started by ohasd is responsible for starting the following processes:
E oraagent
gipcd: The Grid Interprocess Communication (GIPC) daemon is a support process that
enables Redundant Interconnect Usage. Redundant Interconnect Usage enables load
balancing and high availability across multiple (up to four) private networks (also known
as interconnects).
mdnsd: The Multicast Domain Name Service (mDNS) daemon is used by Grid Plug and
Play to locate profiles in the cluster, as well as by GNS to perform name resolution.
evmd: The Event Management daemon (EVM) is a background process that publishes
events that Oracle Clusterware creates.
ASM: Provides disk management for Oracle Clusterware and Oracle Database
gpnpd: Grid Plug and Play (GPNPD) provides access to the Grid Plug and Play profile
and coordinates updates to the profile among the nodes of the cluster to ensure that all
of the nodes have the most recent profile.

Grid Plug and Play
In previous releases, adding or removing servers in a

cluster required extensive manual preparation.

In Oracle Database 11g Release 2, GPnP allows each
node to perform the following tasks dynamically:
Negotiating appropriate network identities for itself
Acquiring additional information from a configuration profile
Configuring or reconfiguring itself using profile data, making a
host names and addresses resolvable on the network has )
om
To add a node, simply connect the server to the cluster
c
s
and allow the cluster to configure the node.nisy uide
b ru nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsadding fe or removing servers in a cluster required extensive manual
e o
With past releases,
rt n-Withtrathe release of Oracle Database 11g Release 2, Grid Plug and Play (GPnP)
v
preparation.
o costs of installing, configuring, and managing server nodes by using a Grid
Ereducesnthe
Naming Service within the cluster to allow each node to perform the following tasks
dynamically:
Negotiating appropriate network identities for itself
Acquiring additional information it needs to operate from a configuration profile
Configuring or reconfiguring itself using profile data, making host names and addresses
resolvable on the network
Because servers perform these tasks dynamically, adding and removing nodes simply
requires an administrator to connect the server to the cluster and to allow the cluster to
configure the node. Using Grid Plug and Play, and using best practices recommendations,
adding a node to the database cluster is part of the normal server restart, and removing a
node happens when a server is turned off. This removes many manual operations, reduces
opportunity for error, and encourages configurations that can be changed more easily than
those requiring fixed per-node configuration.
The best case uses ASM and Automatic Undo Management so there is no particular policy
decision to make if an undo tablespace needs to be allocated for a newly identified database
instance.

Grid Naming Service
GNS is an integral component of Grid Plug and Play.

The only static IP address required for the cluster is the

GNS virtual IP address.
The cluster subdomain is defined as a delegated domain.
[root@my-dns-server ~]# cat /etc/named.conf
// Default initial "Caching Only" name server configuration
...
# Delegate to gns on cluster01
cluster01.example.com #cluster sub-domain# NS cluster01-gns.example.com s a
# Let the world know to go to the GNS vip )ha
m
co
cluster01-gns.example.com 192.0.2.155 #cluster GNS Address
s
A request to resolve cluster01- u n isy uide
scan.cluster01.example.com would be b r nt Gto the
forwarded
n t o@ tude
GNS on 192.0.2.155.
c i me this S
Each node in the clusteraruns
n s asmulticast e DNS (mDNS)
u
process.
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Naming n s fe Service (GNS) assumes that there is a DHCP server running on the
e o
Employing Grid
rtnetwork a enough addresses to assign to the VIPs and single client access name
trwith
v
public n -
o With GNS, only one static IP address is required for the cluster, the GNS virtual
E(SCAN)nVIPs.
IP address. This address should be defined in the DNS domain. GNS sets up a multicast
DNS (mDNS) server within the cluster, which resolves names in the cluster without static
configuration of the DNS server for other node IP addresses.
The mDNS server works as follows: Within GNS, node names are resolved using link-local
multicast name resolution (LLMNR). It does this by translating the LLMNR .local domain
used by the multicast resolution to the subdomain specified in the DNS query. When you
select GNS, an mDNS server is configured on each host in the cluster. LLMNR relies on the
mDNS that Oracle Clusterware manages to resolve names that are being served by that host.
To use GNS, before installation, the DNS administrator must establish domain delegation to
the subdomain for the cluster. Queries to the cluster are sent to the GNS listener on the GNS
virtual IP address. When a request comes to the domain, GNS resolves it using its internal
mDNS and responds to the query.

Single Client Access Name
The single client access name (SCAN) is the address used
by clients connecting to the cluster.
The SCAN is a fully qualified host name located in the GNS

subdomain registered to three IP addresses.
# dig @192.0.2.155 cluster01-scan.cluster01.example.com
...
;; QUESTION SECTION:
;cluster01-scan.cluster01.example.com. IN A
;; ANSWER SECTION:
cluster01-scan.cluster01.example.com. 120 IN A 192.0.2.244 s a
cluster01-scan.cluster01.example.com. 120 IN A 192.0.2.246 )ha
cluster01-scan.cluster01.example.com. 120 IN A 192.0.2.245 m
co
s
isy uide
;; AUTHORITY SECTION:
cluster01.example.com. 10800 IN A 192.0.2.155
u n
r nt G
;; SERVER: 192.0.2.155#53(192.0.2.155)
b
n t o@ t u de
The SCAN provides a stable, highly e is S available name for
clients to use, independent i m
cof thee nodes th that make up the
a s
cluster. o n n o us
e r t e t
v
(e 2012, s
n and/or its affiliates. All rights reserved.
eOracle
n n
Copyright l i c
e r t o
a b le
n Ev access s f er
r t o
The single client
t r a n name (SCAN) is the address used by clients connecting to the
ve The
Ecluster. - is a fully qualified host name (host name + domain) registered to three IP
SCAN
nonIf you use GNS, and you have DHCP support, then the GNS will assign addresses
addresses.
dynamically to the SCAN.
If you do not use GNS, the SCAN should be defined in the DNS to resolve to the three
addresses assigned to that name. This should be done before you install Oracle Grid
Infrastructure. The SCAN and its associated IP addresses provide a stable name for clients to
use for connections, independent of the nodes that make up the cluster.
SCANs function like a cluster alias. However, SCANs are resolved on any node in the cluster,
so unlike a VIP address for a node, clients connecting to the SCAN no longer require updated
VIP addresses as nodes are added to or removed from the cluster. Because the SCAN
addresses resolve to the cluster, rather than to a node address in the cluster, nodes can be
added to or removed from the cluster without affecting the SCAN address configuration.
During installation, listeners are created on each node for the SCAN IP addresses. Oracle
Clusterware routes application requests to the cluster SCAN to the least loaded instance
providing the service.

SCAN listeners can run on any node in the cluster. SCANs provide location independence for
the databases so that client configuration does not have to depend on which nodes run a
particular database.
Oracle Database 11g Release 2 and later instances register with SCAN listeners only as
remote listeners. Upgraded databases register with SCAN listeners as remote listeners, and
also continue to register with all other listeners.
If you specify a GNS domain during installation, the SCAN defaults to clustername-
scan.GNS_domain. If a GNS domain is not specified at installation, the SCAN defaults to

clustername-scan.current_domain.
Note: dig: Domain Information Groper
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Oracle Automatic Storage Management (ASM)
ASM is a volume manager and file system.

ASM operates efficiently in both clustered and

nonclustered environments.
ASM is installed in the Grid Infrastructure home
separate from the Oracle Database home.
Application Application s a
)ha
m
co
File system
s
n sy uide
iASM
Logical Volume Manager b u
r nt G
n t o@ tude
Operating system
me this Operating S system
c i
Hardware n as use Hardware
t o n t o
e v er nse
n ( 2012,
Copyright l i ceOracle and/or its affiliates. All rights reserved.
r n
to able
e
n Ev nStorage
s fer Management (ASM) is a volume manager and file system built in to
Oracle
e r to Database
Automatic
- tra server. Raw disk volumes are allocated to ASM for management and
v
the Oracle n
Econtrolninothe same way that raw volumes are managed by a volume manager. ASM is highly
integrated with, and highly optimized for, the Oracle Database. It has become the best
practice standard for Oracle Database storage.
Combining volume management functions with a file system allows a level of integration and
efficiency that would not otherwise be possible. For example, ASM is able to avoid the
overhead associated with a conventional file system and achieve native raw disk performance
for Oracle data files and other file types supported by ASM.
ASM is engineered to operate efficiently in both clustered and nonclustered environments.
Oracle ASM is installed in the Oracle Grid Infrastructure home separate from the Oracle
Database home.

ASM Key Features and Benefits
Stripes files rather than logical volumes

Provides redundancy on a file basis

Enables online disk reconfiguration and dynamic
rebalancing
Reduces the time significantly to resynchronize a transient
failure by tracking changes while disk is offline
Provides adjustable rebalancing speed s a
) ha
Is cluster-aware o m
c e
Supports reading from mirrored copy insteadisofysprimary id
u n G u
copy for extended clusters
@ br ent
Is automatically installed as part o Grid
enoft the S tudInfrastructure
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E striping
n s feand mirroring without the need to purchase a third-party Logical
e o
ASM provides
tra ASM divides a file into pieces and spreads them evenly across all the
rt Manager.
v
Volume n -
no uses an index technique to track the placement of each piece. Traditional striping
Edisks. ASM
techniques use mathematical functions to stripe complete logical volumes. ASM is unique in
that it applies mirroring on a file basis, rather than on a volume basis. Therefore, the same
disk group can contain a combination of files protected by mirroring or not protected at all.
When your storage capacity changes, ASM does not restripe all the data. However, in an
online operation, ASM moves data proportional to the amount of storage added or removed to
evenly redistribute the files and maintain a balanced I/O load across the disks. You can adjust
the speed of rebalance operations to increase or decrease the speed and adjust the impact
on the I/O subsystem. This capability also enables the fast resynchronization of disks that
may suffer a transient failure.
ASM supports all Oracle database file types. ASM supports Real Application Clusters (RAC)
and eliminates the need for a cluster Logical Volume Manager or a cluster file system. In
extended clusters, you can set a preferred read copy
ASM is included in the Grid Infrastructure installation. It is available for both Enterprise Edition
and Standard Edition installations.

ASM and Grid Infrastructure
ASM provides enterprise-class shared storage for Oracle

RAC databases.
OCR and voting disks can be stored in Oracle ASM.
Storing OCR and the voting disk on Oracle ASM eliminates
the need for third-party cluster volume managers.
Only one Oracle ASM instance is supported on a server.
When managing an ASM instance, the administration s a
h
activity must be performed in the Grid Infrastructure home.
) a
c om
i s ys ide
r u n Gu
@ b ent
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfecan be stored in Oracle Automatic Storage Management (Oracle ASM).
e o
OCR tand votingadisks
r the tr ASM partnership and status table is replicated on multiple disks and is
v
Because n -
Oracle
E no
extended to store OCR, the OCR can tolerate the loss of the same number of disks as are in
the underlying disk group and can be relocated in response to disk failures.
Oracle ASM reserves several blocks at a fixed location on every Oracle ASM disk for storing
the voting disk. If the disk holding the voting disk fails, Oracle ASM selects another disk on
which to store this data.
Storing OCR and the voting disk on Oracle ASM eliminates the need for third-party cluster
volume managers and eliminates the complexity of managing disk partitions for OCR and
voting disks in Oracle Clusterware installations.
Only one Oracle ASM instance is supported on a server. When managing an Oracle ASM
instance, the administration activity must be performed in the Oracle Grid Infrastructure home.

Quiz
The init.ohasd entry in the /etc/inittab file is

responsible for:
a. Starting Oracle Clusterware when the node boots
b. Mounting shared volumes as required by Oracle
Clusterware
c. Managing node evictions
d. Restarting ohasd in the event of a crash s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o d
rt n-tra
v
E answer
The
no is d. The init.ohasd entry in the /etc/inittab file is responsible for
restarting the Oracle High Availability Services daemon (ohasd) in the event of a crash.

Quiz
Which of the following statements regarding Grid Naming

Service is not true?

a. GNS is an integral component of Grid Plug and Play.
b. Each node in the cluster runs a multicast DNS (mDNS)
process.
c. The GNS virtual IP address must be assigned by DHCP.
d. The cluster subdomain is defined as a delegated domain. as a
m )h
s co
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o c
rt n-tra
v
E no c is not correct. The GNS VIP address must be statically defined.
Statement

Quiz
Each cluster nodes public Ethernet adapter must support UDP

or RDS.
a. True
b. False
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o b
rt n-tra
v
This
no is false. Actually, each cluster nodes public Ethernet adapter should support
E statement
TCP/IP. The private adapter should support UDP or RDS on Linux/UNIX platforms and
TCP/IP on Windows platforms.

Summary
In this lesson, you should have learned how to:

Explain the principles and purposes of clusters

Describe the Oracle Clusterware architecture
Describe how Grid Plug and Play affects Clusterware
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Practice 1 Overview
This practice covers installing Oracle Grid Infrastructure.

s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non
RAC Concepts
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives

Explain the necessity of global resources

Describe global cache coordination
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Overview of Oracle RAC
A cluster comprises multiple interconnected servers that

appear as one server to end users and applications.

With Oracle Clusterware, Oracle RAC enables you to
cluster an Oracle database.
Oracle Clusterware enables nonclustered and RAC
databases to use the Oracle high-availability infrastructure.
Oracle Clusterware enables you to create a clustered pool of a
storage to be used by any combination of nonclustered and h a s
)
Oracle RAC databases. om c e
y s
Noncluster Oracle databases have a one-to-one
u n is uid
relationship between the database andbthe
r instance.
n t G
Oracle RAC environments haveen o@ tude
a tone-to-many
m i s S
relationship between theadatabase c i
s se t and instances.h
An Oracle RAC database n n can u have up to 100 instances.
r t o t o
( e ve ense
o n n e2012,
Copyright
v e rt rab
n fe interconnected computers or servers that appear as if they were
E nsmultiple
e o
A cluster comprises
ra users and applications. Oracle RAC enables you to cluster an Oracle
rt nto-tend
v
one server
noOracle RAC uses Oracle Clusterware for the infrastructure to bind multiple servers
Edatabase.
so they operate as a single system.
Oracle Clusterware is a portable cluster management solution that is integrated with Oracle
Database. Oracle Clusterware is also a required component for using Oracle RAC. In
addition, Oracle Clusterware enables both noncluster Oracle databases and Oracle RAC
databases to use the Oracle high-availability infrastructure. Oracle Clusterware enables you
to create a clustered pool of storage to be used by any combination of noncluster and Oracle
RAC databases.
Oracle Clusterware is the only clusterware that you need for most platforms on which Oracle
RAC operates. You can also use clusterware from other vendors if the clusterware is certified
for Oracle RAC.
Noncluster Oracle databases have a one-to-one relationship between the Oracle database
and the instance. Oracle RAC environments, however, have a one-to-many relationship
between the database and instances. An Oracle RAC database can have up to 100
instances, all of which access one database. All database instances must use the same
interconnect, which can also be used by Oracle Clusterware.

Oracle RAC databases differ architecturally from noncluster Oracle databases in that each
Oracle RAC database instance also has:
At least one additional thread of redo for each instance
An instance-specific undo tablespace
The combined processing power of the multiple servers can provide greater throughput and
Oracle RAC scalability than is available from a single server.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

RAC One Node
Single-Instance High Availability
The Oracle RAC One Node option is a single instance of
Oracle RAC running on one node in a cluster.

This option adds to the flexibility that Oracle offers for
consolidation.
Many databases can be consolidated into a single cluster
with minimal overhead while providing:
High-availability benefits of failure protection s a
Online rolling patch application ) ha
o m
Rolling upgrades for the operating system and Oracle
y s c e
s uid
Clusterware. uni b r nt G
Oracle RAC One Node is supported
n t o@ t u de
on all platforms on
which Oracle RAC is certified.
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E 11g s e
fRelease
Oracle
r t o
Database
t r a n 2 introduces a new option, Oracle RAC One Node. Oracle
ve OneoNode
ERAC - is a single instance of Oracle RAC running on one node in a cluster. This
n nto the flexibility that Oracle offers for consolidation. Many databases can be
option adds
consolidated into a single cluster with minimal overhead while providing the high availability
benefits of failure protection, online rolling patch application, as well as rolling upgrades for
the operating system and Oracle Clusterware.
Oracle RAC One Node is supported on all platforms on which Oracle RAC is certified. Oracle
RAC One Node is certified on Oracle Virtual Machine (Oracle VM). Oracle RAC One Node
provides the following benefits:
Always available single-instance database services
Built-in cluster failover for high availability
Live migration of instances across servers
Online rolling patches and rolling upgrades for single-instance databases
Online upgrade from single-instance to multi-instance Oracle RAC
Better consolidation for database servers
Enhanced server virtualization
Lower-cost development and test platform for full Oracle RAC

Oracle RAC One Node
With online database relocation, a RAC One Node instance

can be relocated to another server. This can be useful if:

The current server is running short on resources
The current server requires maintenance operations, such as
operating system patches
The same technique can be used to relocate RAC One
Node instances to high-capacity servers to accommodate a
changes in workload. ha s
m )
Single Client Access Name (SCAN) allows clients o
cto e
y s
connect to the database regardless of where u n isthe service
u id is
located

br ent G
@ udeasily scaled
An Oracle RAC One Node database e nto canS tbe
c is
im ifthconditions
a s
up to a full Oracle RAC databases e demand it.
o n n o u
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
EvRACnOne r
feNode
o n
UsingtOracle a s online database relocation, you can relocate the Oracle RAC
r t r
ve Nodeoninstance
-
EOne n
to another server, if the current server is running short on resources or
requires maintenance operations, such as operating system patches. You can use the same
technique to relocate Oracle RAC One Node instances to high-capacity servers (for example,
to accommodate changes in workload), depending on the resources available in the cluster.
In addition, Resource Manager Instance Caging or memory optimization parameters can be
set dynamically to further optimize the placement of the Oracle RAC One Node instance on
the new server.
Using the Single Client Access Name (SCAN) to connect to the database, clients can locate
the service independently of the node on which it is running. Relocating an Oracle RAC One
Node instance is, therefore, mostly transparent to the client, depending on the client
connection.
With Oracle RAC and Oracle RAC One Node, you can standardize your deployments across
the data center, achieving the required level of scalability and high availability for your
applications. With Oracle RAC One Node, there is no limit to server scalability, and, if
applications grow to require more resources than a single node can supply, then you can
easily scale up your single-instance database online to a full Oracle Real Application Clusters
database.

Oracle RAC One Node and Oracle Clusterware
Being closely related to Oracle RAC, Oracle RAC One

Node requires Oracle Clusterware.

ASM and Oracle Clusterware are installed into a single
home directory,
called Oracle Grid
Infrastructure 11g
Release 2. a
This directory is ha s
m )
referred to as the o
c e
y s
Grid Infrastructure
u n is uid
home. br ent G

n t o@ tud
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EDatabase
n s fe11g Release 2, Automatic Storage Management (ASM) and Oracle
e o
With Oracle
rt n-are trainstalled into a single home directory, collectively called Oracle Grid
v
Clusterware
no This directory is referred to as the Grid Infrastructure home. Configuration
EInfrastructure.
assistants start after the Oracle Universal Installer interview process and binary installation
that configure ASM and Oracle Clusterware.

Cluster-Aware Storage Solutions
RAC databases use a shared-everything

architecture and require cluster-aware

storage for all database files.
The Oracle RAC database software
manages disk access and is certified for
use on a variety of storage architectures.
Oracle Database provides the following file storage options a
for Oracle RAC: ha s
) m
Oracle Automatic Storage Management s co
Oracle Cluster File System (Windows) anduOCFS2n u ide
isy (Linux)
b r nt G
Certified cluster file system or cluster-aware
t o e
@ tudvolume manager
Certified NFS file servers men is S
s c i th
a
n o us e
r t o n t
v e s e
n (e 2012,
r t o n le
e a b
n Ev databases f er
An Oracle
r t o RAC
t r a n is a shared-everything database. All data files, control files, PFILEs,
veredoologn-files in Oracle RAC environments must reside on cluster-aware shared disks, so
Eand
that all n
of the cluster database instances can access these storage components. Because
Oracle RAC databases use a shared-everything architecture, Oracle RAC requires cluster-
aware storage for all database files.
In Oracle RAC, the Oracle Database software manages disk access and is certified for use on
a variety of storage architectures. It is your choice how to configure your storage, but you
must use a supported cluster-aware storage solution. Oracle Database provides the following
file storage options for Oracle RAC:
Oracle Automatic Storage Management (Oracle ASM). Oracle recommends this solution
to manage your storage.
A certified cluster file system, including OCFS2 and Oracle Cluster File System (OCFS
for Windows). OCFS2 is available for Linux, and OCFS for Windows is available for
Windows platforms. However, you may optionally use a third-party cluster file system or
cluster-aware volume manager that is certified for Oracle RAC.
Certified network file system (NFS) file servers

Oracle Cluster File System
Is a shared disk cluster file system for Linux (OCFS2) and

Windows (OCFS)
Provides open solution on the operating system side
OCFS2 can be downloaded from OTN:
http://oss.oracle.com/projects/ocfs2/ (Linux)
OCFS is included in the installation media for Oracle Grid
Infrastructure and Oracle RAC on Windows platforms. s a
ha
It is installed automatically with Oracle Clusterware. m)
s co
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe (OCFS) is a shared file system designed specifically for Oracle
E FilenSystem
Oracle
e t oCluster
rApplicationtraClusters. OCFS eliminates the requirement that Oracle database files be
v
Real n -
Elinked tonological drives and enables all nodes to share a single Oracle Home (on Windows
2000 and 2003 only), instead of requiring each node to have its own local copy. OCFS
volumes can span one shared disk or multiple shared disks for redundancy and performance
enhancements. The following is a list of files that can be placed on Oracle Cluster File System
version 1:
Oracle software installation: Currently, this configuration is supported only on Windows
2000 and 2003. Oracle Cluster File System 2 1.2.1 provides support for Oracle Home on
Linux as well.
Oracle files (control files, data files, redo logs, bfiles, and so on)
Shared configuration files (spfile)
Files created by the Oracle server during run time
Voting and OCR files
Oracle Cluster File System is free for developers and customers. The source code is provided
under the General Public License (GPL) on Linux. It can be downloaded from the Oracle
Technology Network website.
Note: From OTN, you can specifically download OCFS for Linux. However, when you
download the database software for Windows, OCFS is already included.
Benefits of Using RAC
High availability: Surviving node and instance failures

Scalability: Adding more nodes as you need them in the

future
Pay as you grow: Paying for only what you need today
Key grid computing features:
Growth and shrinkage on demand
Single-button addition of servers s a
)ha
Automatic workload management for services m
s co
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EApplication
n s feClusters (RAC) enables high utilization of a cluster of standard, low-
Oracle
e t o
Real
rmodular tra such as blades.
v
cost n -servers
no automatic workload management for services. Services are groups or
ERAC offers
classifications of applications that comprise business components corresponding to
application workloads. Services in RAC enable continuous, uninterrupted database
operations and provide support for multiple services on multiple instances. You assign
services to run on one or more instances, and alternate instances can serve as backup
instances. If a primary instance fails, the Oracle server moves the services from the failed
instance to a surviving alternate instance. The Oracle server also automatically load-balances
connections across instances hosting a service.
RAC harnesses the power of multiple low-cost computers to serve as a single large computer
for database processing, and provides the only viable alternative to large-scale symmetric
multiprocessing (SMP) for all types of applications.
RAC, which is based on a shared-disk architecture, can grow and shrink on demand without
the need to artificially partition data among the servers of your cluster. RAC also offers a
single-button addition of servers to a cluster. Thus, you can easily provide or remove a server
to or from the database.

Clusters and Scalability
SMP model RAC model

Shared
Memory
storage
s a
Cache Cache SGA SGA ) ha
o m
y s c e
CPU CPU CPU CPU BGP BGP un BGPGu
is BGP id
@ br ent
Cache coherency nto Cache d
tufusion
e S
s c im this
n a use BGP (background process)
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
If yourto n E nscales
application s fe transparently on SMP machines, then it is realistic to expect it to
v e
scale r well on
n - tra without having to make any changes to the application code.
RAC,
no the database instance, and the node itself, as a single point of failure, and
ERAC eliminates
ensures database integrity in the case of such failures.
The following are some scalability examples:
Allow more simultaneous batch processes.
Allow larger degrees of parallelism and more parallel executions to occur.
Allow large increases in the number of connected users in online transaction processing
(OLTP) systems.

Levels of Scalability
Hardware: Disk input/output (I/O)

Internode communication: High bandwidth and low latency

Operating system: Number of CPUs
Database management system: Synchronization
Application: Design
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fe of cluster databases requires optimal scalability on four levels:
e t o
Successful implementation
rHardware trascalability: Interconnectivity is the key to hardware scalability, which greatly
v n -
E dependsno on high bandwidth and low latency.
Operating system scalability: Methods of synchronization in the operating system can
determine the scalability of the system. In some cases, potential scalability of the
hardware is lost because of the operating systems inability to handle multiple resource
requests simultaneously.
Database management system scalability: A key factor in parallel architectures is
whether the parallelism is affected internally or by external processes. The answer to
this question affects the synchronization mechanism.
Application scalability: Applications must be specifically designed to be scalable. A
bottleneck occurs in systems in which every session is updating the same data most of
the time. Note that this is not RAC-specific and is true on single-instance systems, too.
It is important to remember that if any of the preceding areas are not scalable (no matter how
scalable the other areas are), then parallel cluster processing may not be successful. A typical
cause for the lack of scalability is one common shared resource that must be accessed often.

This causes the otherwise parallel operations to serialize on this bottleneck. High latency in
the synchronization increases the cost of synchronization, thereby counteracting the benefits
of parallelization. This is a general limitation and not a RAC-specific limitation.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Scaleup and Speedup
Original system
Hardware Time 100% of task
Cluster system scaleup Cluster system speedup
s a
Hardware Time Up to ) ha
200% Hardware o m
of Up to
y s c 100%e
Hardware task 300%
u n is uofidtask
Time of
Hardware b r n t G
task
t o@ tudTime/2 e
n
e is S
Hardware
c i m th
Time
a s e
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n Evability s f er
Scaleup
r t o is the
t r a n sustain the same performance levels (response time) when both
to
ve oand
Eworkload - resources increase proportionally:
n n Scaleup = (volume parallel) / (volume original)
For example, if 30 users consume close to 100 percent of the CPU during normal processing,
then adding more users would cause the system to slow down due to contention for limited
CPU cycles. However, by adding CPUs, you can support extra users without degrading
performance.
Speedup is the effect of applying an increasing number of resources to a fixed amount of
work to achieve a proportional reduction in execution times:
Speedup = (time original) / (time parallel)
Speedup results in resource availability for other tasks. For example, if queries usually take
ten minutes to process, and running in parallel reduces the time to five minutes, then
additional queries can run without introducing the contention that might occur if they were to
run concurrently.

Speedup/Scaleup and Workloads
Workload Speedup Scaleup
OLTP and Internet No Yes
DSS with parallel query Yes Yes

s a
Batch (mixed) Possible Yes
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eworkload s e
fdetermines
The type
r t o of
t r a n whether scaleup or speedup capabilities can be achieved
ve parallel
Eusing o n- processing.
n
Online transaction processing (OLTP) and Internet application environments are
characterized by short transactions that cannot be further broken down and, therefore, no
speedup can be achieved. However, by deploying greater amounts of resources, a larger
volume of transactions can be supported without compromising the response.
Decision support systems (DSS) and parallel query options can attain speedup, as well as
scaleup, because they essentially support large tasks without conflicting demands on
resources. The parallel query capability within the Oracle database can also be leveraged to
decrease overall processing time of long-running queries and to increase the number of such
queries that can be run concurrently.
In an environment with a mixed workload of DSS, OLTP, and reporting applications, scaleup
can be achieved by running different programs on different hardware. Speedup is possible in
a batch environment, but may involve rewriting programs to use the parallel processing
capabilities.

I/O Throughput Balanced: Example
Each machine has 2 CPUs:

2 200 MB/s 4 = 1600 MB/s
Each machine has 2 HBAs:

8 200 MB/s = 1600 MB/s
Each switch needs to support 800 MB/s

to guarantee a total system throughput s a
FC-switch
of 1600 MB/s. )h a
co m
s e
u n isyEach disk
u idarray
Disk Disk Disk Disk Disk Disk Disk brDisk nhas t Gone 2 Gb
array 1 array 2 array 3 array 4 array 5 array 6 array
n t o@7 array t u d8e 8 200 MB/s =
controller:
c i me this S 1600 MB/s
na uss e
n
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe delivers the I/O demand that is required, all system components
E thatnassystem
To make
e o sure
rt I/O path a to be orchestrated to work together.
trneed
v
on the n -
no link determines the I/O throughput.
EThe weakest
On the left, you see a high-level picture of a system. This is a system with four nodes, two
Host Bus Adapters (HBAs) per node, two Fibre Channel switches, which are attached to four
disk arrays each. The components on the I/O path are the HBAs, cables, switches, and disk
arrays. Performance depends on the number and speed of the HBAs, switch speed, controller
quantity, and speed of disks. If any one of these components is undersized, the system
throughput is determined by this component. Assuming you have a 2 Gb HBA, the nodes can
read about 8 200 MB/s = 1.6 GB/s. However, assuming that each disk array has one
controller, all eight arrays can also do 8 200 MB/s = 1.6 GB/s. Therefore, each of the Fibre
Channel switches also needs to deliver at least 2 Gb/s per port, to a total of 800 MB/s
throughput. The two switches will then deliver the needed 1.6 GB/s.
Note: When sizing a system, also take the system limits into consideration. For instance, the
number of bus slots per node is limited and may need to be shared between HBAs and
network cards. In some cases, dual port cards exist if the number of slots is exhausted. The
number of HBAs per node determines the maximal number of Fibre Channel switches. And
the total number of ports on a switch limits the number of HBAs and disk controllers.

Performance of Typical Components
Throughput Performance
Component Theory (Bit/s) Maximal Byte/s

HBA Gb/s 100/200 MB/s
16 Port Switch 8 2 Gb/s 1600 MB/s
Fibre Channel 2 Gb/s 200 MB/s
s a
Disk Controller 2 Gb/s 200 MB/s )ha
co m
GigE NIC 1 Gb/s 80 MB/s sys
u n i u ide
Infiniband 10 Gb/s 890 b nt G
rMB/s
n t o@ tude
CPU
c i me thi200250 sS MB/s
n s
a use

e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E npeoples fe often confuse bits with bytes. This confusion originates mainly from
e o
Whiletdiscussing,
tra vendors tend to describe components performance in bits/s; whereas
r thatnhardware
v
the fact -
novendors and customers describe their performance requirements in bytes/s.
Edatabase
The following is a list of common hardware components with their theoretical performance in
bits/second and typical performance in bytes/second:
HBAs come in 1 or 2 Gb per second with a typical throughput of 100 or 200 MB/s.
A 16 Port Switch comes with sixteen 2 Gb ports. However, the total throughput is 8
times 2 Gb, which results in 1600 MB/s.
Fibre Channel cables have a 2 Gb/s throughput, which translates into 200 MB/s.
Disk Controllers come in 2 Gb/s throughput, which translates into about 200 MB/s.
GigE has a typical performance of about 80 MB/s whereas Infiniband delivers about
160 MB/s.

Necessity of Global Resources
SGA1 SGA2 SGA1 SGA2

1008
1008 1008
1 2
s a
)ha
SGA1 SGA2 SGA1 SGA2
m
co
1009 1008 s
1009
u n isy uide
b r nt G
Lost
n t o@ tude
updates!
1008 c i me th1008 is S
n 4a
s se
o n o u 3
t
er nse t
e v
( 2012,
n n
e r to able
n s fer
Ev nenvironments,
r t o
In single-instance
t r a locking coordinates access to a common resource such as a
vein a table.
Erow n- Locking prevents two processes from changing the same resource (or row) at
the same notime.
In RAC environments, internode synchronization is critical because it maintains proper
coordination between processes on different nodes, preventing them from changing the same
resource at the same time. Internode synchronization guarantees that each instance sees the
most recent version of a block in its buffer cache.
Note: The slide shows you what would happen in the absence of cache coordination. RAC
prevents this problem.

Additional Memory Requirement for RAC
Heuristics for scalability cases:

15% more shared pool

10% more buffer cache
Smaller buffer cache per instance in the case of
single-instance workload distributed across multiple
instances
Current values: s a
)ha
co m
SELECT resource_name, s
current_utilization,max_utilization nisy
u u ide
FROM v$resource_limit b r nt G
WHERE resource_name like 'g%s_%'; to@
n t u de
c i me this S
SELECT * FROM v$sgastat
n s like
aname s e 'KCL%';
WHERE name like 'g_s%' or u
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Ememoryn s feis mostly allocated in the shared pool at SGA creation time. Because
e o
RAC-specific a across instances, you must also account for bigger buffer caches.
rt maynbe-trcached
v
blocks
no when migrating your Oracle Database from single instance to RAC, keeping the
ETherefore,
workload requirements per instance the same as with the single-instance case, about 10%
more buffer cache and 15% more shared pool are needed to run on RAC. These values are
heuristics, based on RAC sizing experience. However, these values are mostly upper bounds.
If you use the recommended automatic memory management feature as a starting point, then
you can reflect these values in your SGA_TARGET initialization parameter.
However, consider that memory requirements per instance are reduced when the same user
population is distributed over multiple nodes.
Actual resource usage can be monitored by querying the CURRENT_UTILIZATION and
MAX_UTILIZATION columns for the Global Cache Services (GCS) and Global Enqueue
Services (GES) entries in the V$RESOURCE_LIMIT view of each instance. You can monitor
the exact RAC memory resource usage of the shared pool by querying V$SGASTAT as shown
in the slide.

Parallel Execution with RAC
Execution slaves have node affinity with the execution

coordinator but will expand if needed.
Node 1 Node 2 Node 3 Node 4
s a
)ha
co m
s
u n isyExecution
u ide
b t G
r ncoordinator
Shared disks n t o@ tude Parallel
c i me this S execution
n s
a use server

e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe
E nsoptimizer
Oracles
r t o cost-based
t r a incorporates parallel execution considerations as a
ve on-component in arriving at optimal execution plans.
Efundamental
In a RAC n environment, intelligent decisions are made with regard to intranode and internode
parallelism. For example, if a particular query requires six query processes to complete the
work and six parallel execution slaves are idle on the local node (the node that the user
connected to), then the query is processed by using only local resources. This demonstrates
efficient intranode parallelism and eliminates the query coordination overhead across multiple
nodes. However, if there are only two parallel execution servers available on the local node,
then those two, and four of another node, are used to process the query. In this manner, both
internode and intranode parallelism are used to speed up query operations.
In real-world decision support applications, queries are not perfectly partitioned across the
various query servers. Therefore, some parallel execution servers complete their processing
and become idle sooner than others. The Oracle parallel execution technology dynamically
detects idle processes and assigns work to these idle processes from the queue tables of the
overloaded processes. In this way, the Oracle server efficiently redistributes the query
workload across all processes. Real Application Clusters further extends these efficiencies to
clusters by enabling the redistribution of work across all the parallel execution slaves of a
cluster.

Summary

Explain the necessity of global resources

Describe global cache coordination
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non
Installing and Configuring Oracle RAC
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives

Install the Oracle database software

Create a cluster database
Perform post-database-creation tasks
Convert a single-instance Oracle database to RAC
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Installing the Oracle Database Software
$ /stage/database/Disk1/runInstaller
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EUniversals e
fInstaller
r t o
The Oracle
t r a n (OUI) is used to install the Oracle Database 11g Release 2
ve software.
(11.2) - Start the OUI by executing the runInstaller command from the root
noofnthe Oracle Database 11g Release 2 CD-ROM or from the software staging
Edirectory
location. You can use the Configure Security Updates window to specify an email address to
receive security updates directly from Oracle Support as they occur. Alternatively, you can
elect to opt out of these alerts. If you want to receive them, supply your email address and
your Oracle Support password, and click Next.
The Download Software Updates page allows you to include database software updates or
patches in the installation. The page allows you to either download them directly or use pre-
downloaded updates. Alternatively, you can choose to bypass software updates entirely.
The Select Installation Option window enables you to create and configure a database, install
database software only, or upgrade an existing database. Select the Install database
software only option and click Next.

s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfeOptions window, select Real Application Clusters database
e o
In thetGrid
r nand Installation
traselect all nodes in your cluster on which the software should be installed. If
v
installation -
o oracle user has not been set up, click SSH Connectivity, then provide the
ESSH fornthe
oracle users password and click Setup. If SSH has already been set up, then click Test.
When finished, click Next to continue. In the Select product languages window, select your
desired languages from the Available Languages list and click the right arrow to promote the
selected languages to the Selected Languages list. Click Next to continue.
In the Select database edition window (not shown in the slide), you select whether to install
the Enterprise Edition or the Standard Edition. Select the Enterprise Edition option and click
Next to continue.

s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Database
n s fe edition page, you can select to install either the Enterprise Edition or
On the
e oSelect
tra Click Next to continue.
rt Edition.
v
Standard n -
no Installation Location window, provide a value for ORACLE_BASE if you have not
EIn the Specify
already done so. The default ORACLE_BASE location is /u01/app/oracle, provided the
RDBMS software is being installed by the oracle account. The Software Location section of
the window enables you to specify a value for the ORACLE_HOME location. The default
ORACLE_HOME location is /u01/app/oracle/product/11.2.0/dbhome_1. Accept the
suggested path or enter your own location. After entering the information, review it for
accuracy, and click the Next button to continue. In the Privileged operating system groups
window, select the operating system group that will act as the OSDBA group. Next, select the
group that will act as the OSOPER group. Click Next to continue.

s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe Checks window verifies the operating system requirements that
E Prerequisite
e o
The Perform
trathe installation to be successful. These requirements include:
t met -for
rbe
v
must n
no operating system check
E Certified
Kernel parameters as required by the Oracle Database software
Required operating system packages and correct revisions
After each successful check, the Status for that check will indicate Succeeded. Any tests that
fail are also reported here. If any tests fail, click the Fix & Check Again button. The Installer
will generate fix-up scripts to correct the system deficiencies if possible. Execute the scripts
as directed by the Installer. The tests will be run again after completing the script executions.
When all tests have succeeded, click the Next button to continue. In the Summary window,
review the Global settings and click Finish.

s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe the OUI will display another window, prompting you to run the
e o
At thetend of theainstallation,
r scripts tr on the nodes you chose for the installation. Follow the instructions to run the
v
root.sh n -
o finished, click the OK button to close the Execute Configuration Scripts window
Escripts.nWhen
and return to the Finish screen. Click Close to complete the installation and close the OUI.

Creating the Cluster Database
s a
) ha
o m
y s c e
u n is uid
$ cd /u01/app/oracle/product/11.2.0/dbhome_1/bin b r n t G
$ ./dbca n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
To create nthe cluster
fe
E nsdatabase, change directory to $ORACLE_HOME/bin on the installing
r t o t r a
ve andoexecute
Enode - the database configuration assistant (DBCA) utility as follows:
n n $ cd /u01/app/oracle/product/11.2.0/dbhome_1/bin
$ ./dbca
The Welcome window appears first. You must select the type of database that you want to
install. Select the Oracle Real Application Clusters (RAC) database option, and then click
Next. The Operations window appears. For a first-time installation, you have only two choices:
the first option enables you to create a database and the other option enables you to manage
database creation templates. Select the Create a Database option, and then click Next to
continue.

Database Type Selection
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Templates
n s fe window appears next. The DBCA tool provides several predefined
e o
The Database
rt types trato choose from, depending on your needs. The templates include:
v
database n -
no Purpose or Transaction Processing
E General
Custom Database
Data Warehouse
In the example in the slide, the General Purpose or Transaction Processing option is
chosen. Click Next to continue.

Database Identification
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe window, you must choose between an administrator-managed
E Identification
e o
In thetDatabase
tra cluster database.
r policy-managed
v
and a n -
no
EAdministrator-managed RAC databases specify a list of cluster nodes where RAC instances
will run. Services may also be specified and associated with preferred and alternative nodes.
There is an explicit association between database services, instances, and cluster nodes.
Policy-based management, a new feature in this release, breaks the explicit association
between services, instances, and cluster nodes. Policy-based management introduces the
concept of server pools, which are logical divisions of a cluster that are dynamically allocated
based on relative importance. Database services are associated with server pools, and RAC
instances are automatically started to satisfy the service to server pool associations. You
specify in which server pool the database resource will run and the number of instances
needed (cardinality). Oracle Clusterware is responsible for placing the database resource on
a server. Server pools are logical divisions of a cluster into pools of servers that are allocated
to host databases or other applications. Server pools are managed using crsctl and
srvctl commands. Names must be unique within the resources defined for the cluster.
You must also choose the global database name and the nodes on which to create the cluster
database. The global database name can be up to 30 characters in length and must begin
with an alphabetical character. When you have finished, click Next to continue.

Cluster Database Management Options
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nOptionss fe window is displayed. For small cluster environments, you may
e o
The Management
tra your cluster with Enterprise Manager Database Control. To do this, select
rt to manage
v
choose n -
no Enterprise Manager check box. If you have Grid Control installed somewhere
Ethe Configure
on your network, you can select the Use Grid Control for Database Management option. If
you select Enterprise Manager with the Grid Control option and the DBCA discovers agents
running on the local node, you can select the preferred agent from a list. Grid Control can
simplify database management in large, enterprise deployments.
You can also configure Database Control to send email notifications when alerts occur. If you
want to configure this, you must supply a Simple Mail Transfer Protocol (SMTP) or outgoing
mail server and an email address. You can also enable daily backups here. You must supply
a backup start time as well as operating system user credentials for this option.
If you want to use Grid Control to manage your database, but have not yet installed and
configured a Grid Control server, do not click either of the management methods. When you
have made your choices, click the Next button to continue.

Passwords for Database Schema Owners
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe window appears next. You must supply passwords for the user
E Credentials
e o
The Database
rt createdtraby the DBCA when configuring your database. You can use the same
v
accounts n -
nofor all of these privileged accounts by selecting the Use the Same Administrative
Epassword
Password for All Accounts option. Enter your password in the Password field, and then enter
it again in the Confirm Password field.
Alternatively, you may choose to set different passwords for the privileged users. To do this,
select the Use Different Administrative Passwords option, enter your password in the
Password field, and then enter it again in the Confirm Password field. Repeat this for each
user listed in the User Name column. Click the Next button to continue.

Database File Locations
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E File n s fe
r o
In thetDatabase
t r a Locations window, you must indicate where the database files are to be
ve You
Estored. - choose your storage type from the drop-down list. Detected (and supported)
oncan
sharedn storage types are available here. You can choose to use a standard template for file
locations, one common location, or Oracle-Managed Files (OMF). This cluster database uses
ASM and Oracle Managed Files. Therefore, select the Use Oracle-Managed Files option, and
enter the disk group name in the Database Area field. Alternatively, you can click the Browse
button to indicate the location where the database files are to be created. When you have
made your choices, click the Next button to continue.

Recovery Configuration
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe
E Configuration
r o
In thetRecovery
t r a window, you can select redo log archiving by selecting Enable
ve oIfnyou
EArchiving. - are using ASM or cluster file system storages, you can also select the fast
n area size in the Recovery Configuration window. The size of the area defaults to
recovery
2048 megabytes, but you can change this figure if it is not suitable for your requirements. If
you are using ASM and a single disk group, the fast recovery area defaults to the ASM Disk
Group. If more than one disk group has been created, you can specify it here. If you use a
cluster file system, the fast recovery area defaults to
$ORACLE_BASE/flash_recovery_area. You may also define your own variables for the
file locations if you plan to use the Database Storage window to define individual file
locations. Select the Enable Archiving check box to enable archiving immediately for the new
cluster database.
When you have completed your entries, click Next, and the Database Content window is
displayed.
Note: For Oracle Database 11g Release 2 (11.2), the flash recovery area has been renamed
fast recovery area. Oracle Enterprise Manager, however, still uses the older vocabulary on
its webpages.

Database Content
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Content
n s fe window, you can choose to install the Sample Schemas included
e o
In thetDatabase
tra distribution. On the Custom Scripts tabbed page, you can choose to run
rthe database
v
with n -
Eyour ownnoscripts as part of the database creation process. When you have finished, click the
Next button to continue to the next window.

Initialization Parameters
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe
E nParameters
r o
In thetInitialization
t r a window, you can set important database parameters. The
ve onare
Eparameters - grouped under four tabs:
n
Memory
Sizing
Character Sets
Connection Mode
On the Memory tabbed page, you can set parameters that deal with memory allocation,
including shared pool, buffer cache, Java pool, large pool, and PGA size. Automatic Memory
Management is the preferred memory management method and can be selected here. On the
Sizing tabbed page, you can adjust the database block size. Note that the default is 8 KB. In
addition, you can set the number of processes that can connect simultaneously to the
database.
By clicking the Character Sets tab, you can change the database character set. You can also
select the default language and the date format. On the Connection Mode tabbed page, you
can choose the connection type that clients use to connect to the database. The default type
is Dedicated Server Mode. If you want to use Oracle Shared Server, click the Shared Server
Mode button. If you want to review the parameters that are not found on the four tabs, click
the All Initialization Parameters button. Click the Use Automatic Memory Management button
and click the Next button to continue.

Database Storage Options
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Storage
n s fewindow provides full control over all aspects of database storage,
e o
The Database
tra data files, and log members. Size, location, and all aspects of extent
rt tablespaces,
v
including n -
no are under your control here.
Emanagement
When you have finished, click the Next button to continue to the next page.

Create the Database
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EOptions s e
fwindow
r t o
The Creation
t r a n appears. You can choose to create the database, or save your
ve session
EDBCA - as a database creation script by selecting the corresponding check box. Select
nonDatabase check box, and then click the Finish button. The DBCA displays the
the Create
Summary screen, giving you a last chance to review all options, parameters, and so on that
have been chosen for your database creation.
Review the summary data. When you are ready to proceed, close the Summary window by
clicking the OK button.

Monitoring Progress
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Monitor s e
fwindow
r t o
The Progress
t r a n appears next. In addition to informing you about how fast the
ve ocreation
Edatabase - is taking place, it also informs you about the specific tasks being performed
n n in real time. When the database creation progress reaches 100 percent, the
by the DBCA
DBCA displays a dialog box announcing the completion of the creation process. It also directs
you to the installation log file location, parameter file location, and Enterprise Manager URL.
By clicking the Password Management button, you can manage the database accounts
created by the DBCA.

Postinstallation Tasks
Download and install the required patch updates.
Verify the cluster database configuration.
$ srvctl config database -d orcl

Database unique name: ORCL
Database name: ORCL
Oracle home: /u01/app/oracle/product/11.2.0/dbhome_1
Oracle user: oracle
Spfile: +DATA/orcl/spfileorcl.ora
Domain: example.com
Start options: open
s a
Stop options: immediate
)ha
Database role: PRIMARY m
co
Management policy: AUTOMATIC
s
Server pools: orcl
u n isy uide
Database instances: orcl1,orcl2
b r nt G
o@ tude
Disk Groups: DATA,FRA
Services: n t
Type: RAC
c i me this S
Database is administrator managed
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E database
n s fe has been successfully created, run the following command to verify
e o
After the cluster
tra Registry configuration in your newly installed RAC environment:
rt nCluster
v
the Oracle -
E no $ srvctl config database -d db_name

Checking Managed Targets
https://host01.example.com:1158/em
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfetask that you should perform if you are using Enterprise Manager
Another
e o postinstallation
rt Controltra or Grid Control is to check whether all the managed nodes and their
v
Database n -
Emanagednoresources are properly registered and available. Open a browser and enter the
address for your Database Control console. Click the Targets tab to verify that all the targets
appear here.
Note: If Enterprise Manager Database Control has not been started on your node, you can
start it with the following commands, logged in as the oracle user, or the account that owns
the database software installation:
$ export ORACLE_UNQNAME=orcl
$ emctl start dbconsole

Background Processes Specific to Oracle RAC
ACMS: Atomic Control File to Memory Service

GTX[0-j]: Global Transaction Process

LMON: Global Enqueue Service Monitor
LMD: Global Enqueue Service Daemon
LMS: Global Cache Service Process
LCK0: Instance Enqueue Process
a
LMHB: Global Cache/Enqueue Service Heartbeat Monitorhas
m )
PING: Interconnect Latency Measurement Process o
c e
y s
RCBG: Result Cache Background Processunis u id

br ent G
RMSn: Oracle RAC Management Processes
@
e n to Stud
RSMN: Remote Slave Monitorim is
a s c e th
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n Ev databases f er
An Oracle
r t o RAC
t r a n has the same processes and memory structures as a single-
ve oOracle
Einstance n- database and additional process and memory structures that are specific to
Oraclen RAC. The global cache service and global enqueue service processes, and the global
resource directory (GRD) collaborate to enable cache fusion. The Oracle RAC processes and
their identifiers are as follows:
Atomic Control File to Memory Service (ACMS): In a RAC environment, the ACMS per-
instance process is an agent that contributes to ensuring a distributed SGA memory
update is either globally committed if successful, or globally aborted if a failure occurs.
Global Transaction Process (GTX[0-j]): The GTX[0-j] processes provide
transparent support for XA global transactions in a RAC environment. The database
automatically tunes the number of these processes based on the workload of XA global
transactions.
Global Enqueue Service Monitor (LMON): The LMON process monitors global
enqueues and resources across the cluster and performs global enqueue recovery
operations.
Global Enqueue Service Daemon (LMD): The LMD process manages incoming remote
resource requests within each instance.

Global Cache Service Process (LMS): The LMS process maintains records of the data
file statuses and each cached block by recording information in the GRD.
The LMS process also controls the flow of messages to remote instances and manages
global data block access and transmits block images between the buffer caches of
different instances. This processing is part of the cache fusion feature.
Instance Enqueue Process (LCK0): The LCK0 process manages noncache fusion
resource requests such as library and row cache requests.
Global Cache/Enqueue Service Heartbeat Monitor (LMHB): LMHB monitors LMON,

LMD, and LMSn processes to ensure they are running normally without blocking or
spinning.
Interconnect Latency Measurement Process (PING): Every few seconds, the process
in one instance sends messages to each instance. The message is received by PING
on the target instance. The time for the round trip is measured and collected.
Result Cache Background Process (RCBG): This process is used for handling
s
invalidation and other messages generated by server processes attached to other
a
) ha
instances in Oracle RAC.
o m

y s c e
Oracle RAC Management Processes (RMSn): The RMSn processes perform
u n is uid
manageability tasks for Oracle RAC. Tasks accomplished by an RMSn process include
b r n t G
creation of resources related to Oracle RAC when new instances are added to the
clusters.
n t o@ tude

c i me this S
Remote Slave Monitor (RSMN): The RSMN process manages background slave
n as use
process creation and communication on remote instances. These background slave
processes perform tasks on behalf of a coordinating process running in another
instance. e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Single Instanceto-RAC Conversion
Single-instance databases can be

converted to RAC using:

DBCA
Enterprise Manager
Single-instance
RCONFIG utility database
DBCA automates most of the
conversion tasks. s a
)ha
Before conversion, ensure that: m
Your hardware and operating s co
RAC database
system are supported u n isy uide
b r nt G
Your cluster nodes have access too@
n t t u de
shared storage
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe Configuration Assistant (DBCA) to convert single-instance Oracle
Ethe Database
You can
e o use
tra The DBCA automates the configuration of the control file attributes,
rt nto-RAC.
v
databases
o undo tablespaces and the redo logs, and makes the initialization parameter file
Ecreatesnthe
entries for cluster-enabled environments. It also configures Oracle Net Services, Oracle
Clusterware resources, and the configuration for RAC database management for use by
Oracle Enterprise Manager or the srvctl utility.
Before you use the DBCA to convert a single-instance database to a RAC database, ensure
that your system meets the conditions:
It is a supported hardware and operating system configuration.
It has shared storage. A supported Cluster File System, NFS mount, or ASM is available
and accessible from all nodes.
Your applications have no design characteristics that preclude their use with cluster
database processing.
You can also use Enterprise Manager and the rconfig utility to perform the single instance
to-RAC conversion.

Considerations for Converting Single-Instance
Databases to Oracle RAC
Backup procedures should be available before conversion
takes place.
Archiving in Oracle RAC environments requires a thread
number in the archive file format.
The archived logs from all instances of an Oracle RAC
database are required for media recovery.
By default, all database files are migrated to Oracle s a
Managed Files (OMF). ) ha
c om
i s ys ide
r u n Gu
@ b ent
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nyou s fechoose to use for the conversion, note the following administrative
e o
Whatever method
a converting single-instance databases to Oracle RAC:
rt n-trbefore
v
considerations
E no
Backup procedures should be available before converting from a single-instance Oracle
Database to Oracle RAC. For archiving with Oracle RAC environments, the archive file
format requires a thread number.
The archived logs from all instances of an Oracle RAC database are required for media
recovery. Because of this, if you archive to a file and you do not use a cluster file
system, or some other means to provide shared file systems, then you require a method
of accessing the archive logs from all nodes on which the cluster database has
instances.
By default, all database files are migrated to Oracle Managed Files (OMF). This feature
simplifies tablespace creation, ensures data file location consistency and compliance
with OFA rules, and reduces human error with data file management.

Single-Instance Conversion Using the DBCA
Conversion steps for a single-instance database on

nonclustered hardware:
1. Back up the original single-instance database.
2. Complete the Oracle Grid Infrastructure installation.
3. Validate the cluster.
4. Copy the preconfigured database image.
5. Install the Oracle Database 11g software with RAC. s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe
E ansingle-instance
r t o
To convert from
t r a Oracle database that is on a noncluster computer to a RAC
ve operform
Edatabase, n- the following steps:
n
2. Complete the Oracle Grid Infrastructure installation.
3. Validate the cluster.
4. Copy the preconfigured database image.
5. Install the Oracle Database 11g software with RAC.

Conversion Steps
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Ethe Original
n s fe Single-Instance Database.
1. Back
e t
rtheo Up
trtoacreate a preconfigured image of your single-instance database by using the
v
Use n
DBCA -
Efollowingnoprocedure:
1. Navigate to the bin directory in $ORACLE_HOME, and start the DBCA.
2. In the Welcome window, click Next.
3. In the Operations window, select Manage Templates, and click Next.
4. In the Template Management window, select Create a database template and From
an existing database (structure as well as data), and click Next.
5. In the Source Database window, enter the database name in the Database instance
field, and click Next.
6. In the Template Properties window, enter a template name in the Name field. By default,
the template files are generated in $ORACLE_HOME/assistants/dbca/templates.
Enter a description of the file in the Description field, and change the template file
location in the Template data file field if you want. When you have finished, click Next.
7. In the Location of Database Related Files window, select Maintain the file locations, so
that you can restore the database to the current directory structure, and click Finish. The
DBCA generates two files: a database structure file (template_name.ctl) and a
database preconfigured image file (template_name.dfb).

Conversion Steps
2. Perform the preinstallation steps.

Tasks include kernel parameter configuration, hardware

setup, network configuration, and shared storage setup.
3. Set up and validate the cluster.
Create a cluster with the required number of nodes
according to your hardware vendors documentation.
Validate cluster components before installation.
s a
Install Oracle Clusterware.
) ha
om
Validate the completed cluster installation by using cluvfy.
c e
4. Copy the preconfigured database image. nis y s
u u id
The database structure *.ctl file
br ent G
@ d
The preconfigured database image
e nto*.dfb S tufile
5. Install the Oracle Database s c im Release
11g e t his 2 software with
RAC. n na us
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe
Ethe Preinstallation
r t o
2. Perform
t r a Steps.
ve othe
EComplete - Oracle Clusterware installation, as described in the Oracle Grid Infrastructure
n nGuide for your platform.
Installation
3. Set Up and Validate the Cluster.
Form a cluster with the required number of nodes according to your business needs. When
you have configured all the nodes in your cluster, validate cluster components by using the
Cluster Verification Utility, and then install Oracle Clusterware. When the clusterware is
installed, validate the completed cluster installation and configuration by using the Cluster
Verification Utility.
4. Copy the Preconfigured Database Image.
This includes copying the database structure *.ctl file and the database preconfigured
image *.dfb file that the DBCA created in step one (Back Up the Original Single-Instance
Database) to a temporary location on the node in the cluster from which you plan to run the
DBCA.

5. Install the Oracle Database 11g Release 2 Software with RAC.
1. Run the OUI to perform an Oracle database installation with RAC. Select Cluster
Installation Mode and select the nodes to include in your RAC database.
2. In the OUI Database Configuration Types window, select Advanced install. After
installing the software, the OUI runs postinstallation tools such as NETCA, DBCA, and
so on.
3. In the DBCA Template Selection window, use the template that you copied to a
temporary location in the Copy the Preconfigured Database Image step. Use the
Browse option to select the template location.
4. After creating the RAC database, the DBCA displays the Password Management
window in which you must change the passwords for database privileged users. When
the DBCA exits, the conversion is complete.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Single-Instance Conversion Using rconfig
1. Locate the appropriate .xml file located in the

$ORACLE_HOME/assistants/rconfig/sampleXMLs
directory.
2. Modify the ConvertToRAC_AdminManaged.xml or
ConvertToRAC_PolicyManaged.xml file as required
for your system.
3. Save the file under a different name. a
a s
m )h
$ cd $ORACLE_HOME/assistants/rconfig/sampleXMLs
s co
$ vi ConvertToRAC_PolicyManaged.xml
u n isy uide
... Saved as my_rac_conversion.xml b r nt G
$ rconfig my_rac_conversion.xmlto@ tude
i m en is S
a s c e th
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n
v
Ethe s f er command-line utility to convert a single-instance database to RAC.
You can
Toe r
use
use
- t r an perform the following steps:
tothis feature,rconfig
Ev1. Gonotonthe $ORACLE_HOME/assistants/rconfig/sampleXMLs directory as the

oracle user and open the ConvertToRAC_AdminManaged.xml or the
ConvertToRAC_PolicyManaged.xml file (depending on your desired management
style) using a text editor, such as vi.
2. Review the XML file, and modify the parameters as required for your system. The XML
sample file contains comment lines that provide instructions about how to configure the
file. When you have finished making changes, save the file with the syntax
filename.xml. Make a note of the name you select.
3. Assuming that you save your XML file as my_rac_conversion.xml, navigate to the
$ORACLE_HOME/bin directory, and use the following syntax to run the rconfig
command:
$ ./rconfig my_rac_conversion.xml
Note: The Convert verify option in the .xml file has three options:
Convert verify="YES": rconfig performs checks to ensure that the prerequisites
for single-instance to RAC conversion have been met before it starts conversion.

Convert verify="NO": rconfig does not perform prerequisite checks, and starts
conversion.
Convert verify="ONLY": rconfig performs only prerequisite checks; it does not
start conversion after completing prerequisite checks.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Quiz
The RAC database software installation is initiated by executing

runInstaller from the root directory of the Oracle Database

11g Release 2 CD-ROM or from the software staging location.
1. True
2. False
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o a
rt n-tra
v
no is true.
E statement
The

Quiz
A single-instance database can be converted to a RAC

database by using (choose the correct options):

a. rconfig
b. netca
c. dbca
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o a, c
rt n-tra
v
EChoices
noa and c are correct.

Summary

Install the Oracle database software

Create a cluster database
Perform post-database-creation tasks
Convert a single-instance Oracle database to RAC
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Practice 3 Overview
This practice covers the following topics:

Installing the Oracle database software

Creating a RAC database
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non
Oracle RAC Administration
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives

Use Enterprise Manager Cluster Database pages

Define redo log files in a RAC environment
Define undo tablespaces in a RAC environment
Start and stop RAC databases and instances
Modify initialization parameters in a RAC environment
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Cluster Database Home Page
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EDatabase
n s feHome page serves as a crossroad for managing and monitoring all
e o
The Cluster
traRAC database. From this page, you can access the other main cluster
rt of nyour
v
aspects -
notabs: Performance, Availability, Server, Schema, Data Movement, Software and
Edatabase
Support, and Topology.
On this page, you find General, High Availability, Space Summary, and Diagnostic Summary
sections for information that pertains to your cluster database as a whole. The number of
instances is displayed for the RAC database, in addition to the status. A RAC database is
considered to be up if at least one instance has the database open. You can access the
Cluster Home page by clicking the Cluster link in the General section of the page.
Other items of interest include the date of the last RMAN backup, archiving information, space
utilization, and an alert summary. By clicking the link next to the Flashback Database Logging
label, you can open the Recovery Settings page from where you can change various recovery
parameters.
The Alerts table shows all the recent alerts that are open. Click the alert message in the
Message column for more information about the alert. When an alert is triggered, the name of
the metric for which the alert was triggered is displayed in the Name column.

Cluster Database Home Page
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe provides information about alerts for related targets, such as
EAlertsntable
e o
The Related
rt and ra and contains details about the message, the time the alert was triggered,
tHosts,
v
Listeners n -
noand the time the alert was last checked.
Ethe value,
The Policy Trend Overview page (accessed by clicking the Compliance Score link) provides a
comprehensive view about a group or targets containing other targets with regard to
compliance over a period of time. Using the tables and graphs, you can easily watch for
trends in progress and changes.
The Security At a Glance page shows an overview of the security health of the enterprise for
the targets or specific groups. This helps you to focus on security issues by showing statistics
about security policy violations and noting critical security patches that have not been applied.
The Job Activity table displays a report of the job executions that shows the scheduled,
running, suspended, and problem (stopped/failed) executions for all Enterprise Manager jobs
on the cluster database.
The Instances table lists the instances for the cluster database, their availability, alerts, policy
violations, performance findings, and related ASM Instance. Click an instance name to go to
the Home page for that instance. Click the links in the table to get more information about a
particular alert, advice, or metric.

Cluster Database Instance Home Page
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EDatabase
n s feInstance Home page enables you to view the current state of the
e o
The Cluster
tra a series of metrics that portray its overall health. This page provides a
rt byndisplaying
v
instance -
o for the performance, administration, and maintenance of the instance
Elaunchnpoint
environment.
You can access the Cluster Database Instance Home page by clicking one of the instance
names from the Instances section of the Cluster Database Home page. This page has
basically the same sections as the Cluster Database Home page.
The difference is that tasks and monitored activities from these pages apply primarily to a
specific instance. For example, clicking the Shutdown button on this page shuts down only
this one instance. However, clicking the Shutdown button on the Cluster Database Home
page gives you the option of shutting down all or specific instances.
By scrolling down on this page, you see the Alerts, Related Alerts, Policy Violations, Jobs
Activity, and Related Links sections. These provide information similar to that provided in the
same sections on the Cluster Database Home page.

Cluster Home Page
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E youn s fe Cluster Home page, which can be accessed by clicking the Cluster
e o
The slide shows
rt nin-tthe the
ratop-right corner of Enterprise Manager. Even if the database is down, the
v
tab located
o is available to manage resources. The cluster is represented as a composite
EClusternpage
target composed of nodes and cluster databases. An overall summary of the cluster is
provided here. The Cluster Home page displays several sections, including General,
Configuration, Diagnostic Summary, Cluster Databases, Alerts, and Hosts. The General
section provides a quick view of the status of the cluster, providing basic information such as
current Status, Availability, Up nodes, Clusterware Version, and Oracle Home.
The Configuration section enables you to view the operating systems (including hosts and OS
patches) and hardware (including hardware configuration and hosts) for the cluster.
The Cluster Databases table displays the cluster databases (optionally associated with
corresponding services) associated with this cluster, their availability, and any alerts on those
databases. The Alerts table provides information about any alerts that have been issued
along with the severity rating of each.
The Hosts table (not shown in the screenshot) displays the hosts for the cluster, availability,
corresponding alerts, CPU and memory utilization percentage, and total input/output (I/O) per
second.

Configuration Section
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EHomenpage
s fe is invaluable for locating configuration-specific data. Locate the
e o
The Cluster
ra on the Cluster Home page. The View drop-down list enables you to
rt n-tsection
v
Configuration
o
Einspectnhardware and operating system overview information.
Click the Hosts link, and then click the Hardware Details link of the host that you want. On the
Hardware Details page, you find detailed information about your CPU, disk controllers,
network adapters, and so on. This information can be very useful when determining the Linux
patches for your platform.
Click History to access the hardware history information for the host.
Some hardware information is not available, depending on the hardware platform.
Note: The Local Disk Capacity (GB) field shows the disk space that is physically attached
(local) to the host. This value does not include disk space that may be available to the host
through networked file systems.

Configuration Section
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E System
n s feDetails General page displays the operating system details for a host,
e o
The Operating
rt n-tra
v
including:
no information, such as the distributor version and the maximum swap space of
E General
the operating system
Information about operating system properties
The Source column displays where Enterprise Manager obtained the value for each operating
system property.
To see a list of changes to the operating system properties, click History.
The Operating System Details File Systems page displays information about one or more file
systems for the selected hosts:
Name of the file system on the host
Type of mounted file systemfor example, ufs or nfs
Directory where the file system is mounted
Mount options for the file systemfor example, ro, nosuid, or nobrowse
The Operating System Details Packages page displays information about the operating
system packages that have been installed on a host.

Topology Viewer
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EEnterprise
n s feManager Topology Viewer enables you to visually see the relationships
e o
The Oracle
rt target a for each host of your cluster database. You can zoom in or out, pan, and
trtypes
v
between n -
no details. These views can also be used to launch various administration
Esee selection
functions.
The Topology Viewer populates icons on the basis of your system configuration. If a listener is
serving an instance, a line connects the listener icon and the instance icon. Possible target
types are:
Interface
Listener
ASM Instance
Database Instance
If the Show Configuration Details option is not selected, the topology shows the monitoring
view of the environment, which includes general information such as alerts and overall status.
If you select the Show Configuration Details option, additional details are shown in the
Selection Details window, which are valid for any topology view. For example, the Listener
component would also show the machine name and port number.
You can click an icon and then right-click to display a menu of available actions.

Enterprise Manager Alerts and RAC
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EEnterprise
n s feManager to administer alerts for RAC environments. Enterprise
You can
e o use
tra between database- and instance-level alerts in RAC environments.
rt distinguishes
v
Manager n -
no Manager also responds to metrics from across the entire RAC database and
EEnterprise
publishes alerts when thresholds are exceeded. Enterprise Manager interprets both
predefined and customized metrics. You can also copy customized metrics from one cluster
database instance to another, or from one RAC database to another. A recent alert summary
can be found on the Database Control Home page. Notice that alerts are sorted by relative
time and target name.

Enterprise Manager Metrics and RAC
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe
E forninstance-level
r t o
Alert thresholds
t r a alerts, such as archive log alerts, can be set at the instance
ve level.
Etarget - enables you to receive alerts for the specific instance if performance
nThis
exceeds noyour threshold. You can also configure alerts at the database level, such as setting
alerts for tablespaces. This enables you to avoid receiving duplicate alerts at each instance.

Enterprise Manager Metrics and RAC
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E ntosview fe the metric across the cluster in a comparative or overlay fashion. To
e o
It is also
rthis possible
tra click the Compare Targets link at the bottom of the corresponding metric
t information,
v
view n -
no the Compare Targets page appears, choose the instance targets that you want
Epage. When
to compare by selecting them and then clicking the Move button. If you want to compare the
metric data from all targets, click the Move All button. After making your selections, click the
OK button to continue.
The Metric summary page appears next. Depending on your needs, you can accept the
default timeline of 24 hours or select a more suitable value from the View Data drop-down list.
If you want to add a comment regarding the event for future reference, enter a comment in the
Comment for Most Recent Alert field, and then click the Add Comment button.

Enterprise Manager Alert History and RAC
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfeyou can see a summary of the alert history for each participating
e o
In a RAC environment,
rt directlytrafrom the Cluster Database Home page. The drilldown process is shown in
v
instance n -
noYou click the Alert History link in the Related Links section of the Cluster Database
Ethe slide.
Home page. This takes you to the Alert History page on which you can see the summary for
both instances in the example. Click the Alert History chart for the instance to go directly to
the Alert History page for that instance. From there, you can access a corresponding alert
page by choosing the alert of your choice.

Enterprise Manager Blackouts and RAC
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EEnterprise
n s feManager to define blackouts for all managed targets of your RAC
You can
e o use
tra alerts from being recorded. Blackouts are useful when performing
rt tonprevent
v
database -
no or unscheduled maintenance or other tasks that might trigger extraneous or
Escheduled
unwanted events. You can define blackouts for an entire cluster database or for specific
cluster database instances.
To create a blackout event, click the Setup link at the top of any Enterprise Manager page.
Then click the Blackouts link on the left. The Blackouts page appears.
Click the Create button. The Create Blackout: Properties page appears. You must enter a
name or tag in the Name field. If you want, you can also enter a descriptive comment in the
Comments field. This is optional. Enter a reason for the blackout in the Reason field.
In the Targets area of the Properties page, you must choose a target Type from the drop-
down list. In the example in the slide, the entire cluster database RDBB is chosen. Click the
cluster database in the Available Targets list, and then click the Move button to move your
choice to the Selected Targets list. Click the Next button to continue.
The Member Targets page appears next. Expand the Selected Composite Targets tree and
ensure that all targets that must be included appear in the list. Continue and define your
schedule as you normally would.

Redo Log Files and RAC
Node1 Node2
RAC01 RAC02
Shared storage
Group 1 SPFILE
Group 4
s a
Group 2 RAC01.THREAD=1
RAC02.THREAD=2 Group 5
)ha
co m
Group 3
s
Thread 1 u n isy Thread
u ide2
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe of a server pool in a policy-managed database and a new
E thencardinality
e o
If youtincrease
tra to the server pool, then Oracle Clusterware starts an instance on the new
r is allocated
v
server n -
o have Oracle Managed Files (OMF) enabled. If the instance starts and there is no
Eserver nif you
thread or redo log file available, then Oracle Clusterware automatically enables a thread of
redo and allocates the associated redo log files and undo if the database uses Oracle ASM or
any cluster file system.
You should create redo log groups only if you are using administrator-managed databases.
For administrator-managed databases, each instance has its own online redo log groups.
Create these redo log groups and establish group members. To add a redo log group to a
specific instance, specify the INSTANCE clause in the ALTER DATABASE ADD LOGFILE
statement. If you do not specify the instance when adding the redo log group, the redo log
group is added to the instance to which you are currently connected. Each instance must
have at least two groups of redo log files. You must allocate the redo log groups before
enabling a new instance with the ALTER DATABASE ENABLE INSTANCE instance_name
command. When the current group fills, an instance begins writing to the next log file group. If
your database is in ARCHIVELOG mode, each instance must save full online log groups as
archived redo log files that are tracked in the control file.
Note: You can use Enterprise Manager to administer redo log groups in a RAC environment.

Automatic Undo Management and RAC
Pending Node1 Node2

offline
RAC01 RAC02
Consistent reads
Transaction recovery
Shared storage
s a
SPFILE
) ha
undotbs3 m
RAC01.UNDO_TABLESPACE=undotbs1 o
c e
y s
undotbs1
RAC02.UNDO_TABLESPACE=undotbs2
u n is uid
undotbs2

b r n t G
n t o@ tude
c i me this S
ALTER SYSTEM SET UNDO_TABLESPACE=undotbs3
n as use SID='RAC01';
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Edatabase s e
fautomatically
r t o
The Oracle
t r a n manages undo segments within a specific undo
ve onthat
Etablespace - is assigned to an instance. Under normal circumstances, only the instance
assigned n to the undo tablespace can modify the contents of that tablespace. However, all
instances can always read all undo blocks for consistent-read purposes. Also, any instance
can update any undo tablespace during transaction recovery, as long as that undo tablespace
is not currently used by another instance for undo generation or transaction recovery.
You assign undo tablespaces in your RAC database by specifying a different value for the
UNDO_TABLESPACE parameter for each instance in your SPFILE or individual PFILEs. If you do
not set the UNDO_TABLESPACE parameter, each instance uses the first available undo
tablespace. For policy-managed databases, Oracle automatically allocates the undo
tablespace when the instance starts if you have OMF enabled.
You can dynamically switch undo tablespace assignments by executing the ALTER SYSTEM
SET UNDO_TABLESPACE statement. You can run this command from any instance. In the
example in the slide, the previously used undo tablespace assigned to the RAC01 instance
remains assigned to it until RAC01s last active transaction commits. The pending offline
tablespace may be unavailable for other instances until all transactions against that
tablespace are committed. You cannot simultaneously use Automatic Undo Management
(AUM) and manual undo management in a RAC database. It is highly recommended that you
use the AUM mode.

Starting and Stopping RAC Instances
Multiple instances can open the same database

simultaneously.
Shutting down one instance does not interfere with other
running instances.
SHUTDOWN TRANSACTIONAL LOCAL does not wait for
other instances transactions to finish.
RAC instances can be started and stopped by using: s a
) ha
Enterprise Manager m
The Server Control (srvctl) utility
o
c e
y s
SQL*Plus u n is uid
b r n t G
Shutting down a RAC database means n t o@ shutting t u de down all
instances accessing the database. c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfemultiple instances can have the same RAC database open at the
In a RAC
e o environment,
trashutting down one instance does not interfere with the operation of other
rttime.nAlso,
v
same
o -
Erunningninstances.
The procedures for starting up and shutting down RAC instances are identical to the
procedures used in single-instance Oracle, with the following exception:
The SHUTDOWN TRANSACTIONAL command with the LOCAL option is useful to shut down an
instance after all active transactions on the instance have either committed or rolled back.
Transactions on other instances do not block this operation. If you omit the LOCAL option, this
operation waits until transactions on all other instances that started before the shutdown are
issued either a COMMIT or a ROLLBACK.
You can start up and shut down instances by using Enterprise Manager, SQL*Plus, or Server
Control (srvctl). Both Enterprise Manager and srvctl provide options to start up and shut
down all the instances of a RAC database with a single step.
Shutting down a RAC database mounted or opened by multiple instances means that you
need to shut down every instance accessing that RAC database. However, having only one
instance opening the RAC database is enough to declare the RAC database open.

Starting and Stopping
RAC Instances with srvctl
start/stop syntax:
srvctl start|stop instance -d <db_name> -i <inst_name_list>
[-o open|mount|nomount|normal|transactional|immediate|abort>]
srvctl start|stop database -d <db_name>

[-o open|mount|nomount|normal|transactional|immediate|abort>]
Examples: s a
)ha
m
co
$ srvctl start instance -d orcl -i orcl1,orcl2
s
$ srvctl stop instance -d orcl -i orcl1,orcl2
u n isy uide
b r nt G
$ srvctl start database -d orcl -o openo@
n t t u de
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
The srvctln fe
Estartnsdatabase command starts a cluster database, its enabled instances, and
r t o t r a
ve oThe
Eservices. - srvctl stop database command stops a database, its instances, and its
n n
services.
The srvctl start instance command starts instances of a cluster database. This command
also starts all enabled and nonrunning services that have the listed instances either as
preferred or as available instances.
The srvctl stop instance command stops instances, and all enabled and running services
that have these instances as either preferred or available instances.
You must disable an object that you intend to keep stopped after you issue a srvctl stop
command; otherwise, Oracle Clusterware can restart it as a result of another planned
operation. srvctl does not support concurrent executions of commands on the same object.
Therefore, run only one srvctl command at a time for each database, service, or other
object. In order to use the START or STOP options of the SRVCTL command, your service must
be an Oracle Clusterwareenabled, nonrunning service.
Note: For more information, refer to the Oracle Clusterware and Oracle Real Application
Clusters Administration and Deployment Guide.

Starting and Stopping
RAC Instances with SQL*Plus
[host01] $ echo $ORACLE_SID
orcl1
sqlplus / as sysdba
SQL> startup
SQL> shutdown immediate
[host02] $ echo $ORACLE_SID

orcl2
sqlplus / as sysdba
s a
SQL> startup
)ha
SQL> shutdown immediate m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fore shut down just one instance, and you are connected to your local
e o
If youtwant to start
r thennyou up
a first ensure that your current environment includes the system identifier
trmust
v
node, -
E(SID) fornothe local instance.
To start up or shut down your local instance, initiate a SQL*Plus session connected as
SYSDBA or SYSOPER, and then issue the required command (for example, STARTUP).
You can start multiple instances from a single SQL*Plus session on one node by way of
Oracle Net Services. To achieve this, you must connect to each instance by using a Net
Services connection string, typically an instance-specific alias from your tnsnames.ora file.
For example, you can use a SQL*Plus session on a local node to shut down two instances on
remote nodes by connecting to each using the instances individual alias name.
It is not possible to start up or shut down more than one instance at a time in SQL*Plus, so
you cannot start or stop all the instances for a cluster database with a single SQL*Plus
command.
To verify that instances are running, on any node, look at V$ACTIVE_INSTANCES.
Note: SQL*Plus is integrated with Oracle Clusterware to make sure that corresponding
resources are correctly handled when starting up and shutting down instances via SQL*Plus.

Switch Between Automatic
and Manual Policies
$ srvctl config database -d orcl -a
Database unique name: orcl

Database name: orcl
Oracle user: oracle
Domain:
Start options: open
Database role: PRIMARY
s a
)ha
Server pools: orcl m
co
s
isy uide
Database instances: orcl1,orcl2
Disk Groups: DATA, FRA
u n
r nt G
Services: b
Database is enabled
n t o@ tude
c i me this S
s se
naorcl
srvctl modify database n-d
o o u-y MANUAL;
t
er nse t
e v
( 2012,
n n
e r to able
n
By default, Oracle
fer
Ev nClusterware
s is configured to start the VIP, listener, instance, ASM,
r t o t r a
ve oton-have their profile parameterduring
database
Eresources
services, and other resources system boot. It is possible to modify some
n AUTO_START set to the value 2. This means that after
node reboot, or when Oracle Clusterware is started, resources with AUTO_START=2 need to be
started manually via srvctl. This is designed to assist in troubleshooting and system
maintenance. When changing resource profiles through srvctl, the command tool
automatically modifies the profile attributes of other dependent resources given the current
prebuilt dependencies. The command to accomplish this is:
srvctl modify database -d <dbname> -y AUTOMATIC|MANUAL
To implement Oracle Clusterware and Real Application Clusters, it is best to have Oracle
Clusterware start the defined Oracle Clusterware resources during system boot, which is the
default. The first example in the slide uses the srvctl config database command to
display the current policy for the orcl database. As you can see, it is currently set to its
default: AUTOMATIC. The second statement uses the srvctl modify database command to
change the current policy to MANUAL for the orcl database. When you add a new database by
using the srvctl add database command, by default, that database is placed under the
control of Oracle Clusterware using the AUTOMATIC policy. However, you can use the srvctl
command directly to set the policy to MANUAL: srvctl add database -d orcl -y MANUAL.
Note: You can also use this procedure to configure your system to prevent Oracle
Clusterware from auto-restarting failed database instances more than once.

RAC Initialization Parameter Files
An SPFILE is created if you use the DBCA.

The SPFILE must be created in an ASM disk group or a

cluster file system file.
All instances use the same SPFILE.
If the database is created manually, create an SPFILE
from a PFILE.
s a
)ha
Node1 Node2 m
s co
RAC01
u n isy uide
RAC02
initRAC01.ora
b n G
r initRAC02.ora
t
SPFILE= t o@ tudSPFILE= e
n
SPFILE
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
When you n E nsfdatabase,
create the
e DBCA creates an SPFILE in the file location that you specify.
r t o t r a
ve location
EThis -can be an ASM disk group or a cluster file system file. If you manually create
non it is recommended to create an SPFILE from a PFILE.
your database,
All instances in the cluster database use the same SPFILE. Because the SPFILE is a binary
file, do not edit it. Change the SPFILE settings by using EM or ALTER SYSTEM SQL statements.
RAC uses a traditional PFILE only if an SPFILE does not exist or if you specify PFILE in your
STARTUP command. Using SPFILE simplifies administration, maintaining parameter settings
consistent, and guarantees parameter settings persistence across database shutdown and
startup. In addition, you can configure RMAN to back up your SPFILE.

SPFILE Parameter Values and RAC
You can change parameter settings using the ALTER

SYSTEM SET command from any instance:

ALTER SYSTEM SET <dpname> SCOPE=MEMORY sid='<sid|*>';
SPFILE entries such as:
*.<pname> apply to all instances
<sid>.<pname> apply only to <sid>
<sid>.<pname> takes precedence over *.<pname> s a
) ha
Use current or future *.<dpname> settings for <sid>: m
o
c e
y s
ALTER SYSTEM RESET <dpname> SCOPE=MEMORY sid='<sid>';
u n is uid
b r n t G
Remove an entry from your SPFILE:
n t o@ tude
i m e is S
ALTER SYSTEM RESET <dpname>s SCOPE=SPFILE
a c e th sid='<sid|*>';
n n u s
r t o t o
( e ve ense
o n n e2012,
Copyright
v e rt rab
You can n E nsvalue
modify the
fe of your initialization parameters by using the ALTER SYSTEM SET
e o
rt nThis trais the same as with a single-instance database except that you have the
v
command. -
noto specify the SID clause in addition to the SCOPE clause.
Epossibility
By using the SID clause, you can specify the SID of the instance where the value takes effect.
Specify SID='*' if you want to change the value of the parameter for all instances. Specify
SID='sid' if you want to change the value of the parameter only for the instance sid. This
setting takes precedence over previous and subsequent ALTER SYSTEM SET statements that
specify SID='*'. If the instances are started up with an SPFILE, then SID='*' is the default if
you do not specify the SID clause.
If you specify an instance other than the current instance, then a message is sent to that
instance to change the parameter value in its memory if you are not using the SPFILE scope.
The combination of SCOPE=MEMORY and SID='sid' of the ALTER SYSTEM RESET command
allows you to override the precedence of a currently used <sid>.<dparam> entry. This allows
for the current *.<dparam> entry to be used, or for the next created *.<dparam> entry to be
taken into account on that particular SID.
Using the last example, you can remove a line from your SPFILE.
Note: When you start an instance with an SPFILE, the default for SQL*Plus is SCOPE=BOTH.

EM and SPFILE Parameter Values
SCOPE=MEMORY
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E the n s fe
You can
r t o access
t r a Initialization Parameters page by clicking the Initialization Parameters link
vethe Cluster
Eon o n- Database Server page.
n
The Current tabbed page shows you the values currently used by the initialization parameters
of all the instances accessing the RAC database. You can filter the Initialization Parameters
page to show only those parameters that meet the criteria of the filter that you entered in the
Name field.
The Instance column shows the instances for which the parameter has the value listed in the
table. An asterisk (*) indicates that the parameter has the same value for all remaining
instances of the cluster database.
Choose a parameter from the Select column and perform one of the following steps:
Click Add to add the selected parameter to a different instance. Enter a new instance
name and value in the newly created row in the table.
Click Reset to reset the value of the selected parameter. Note that you can reset only
those parameters that do not have an asterisk in the Instance column. The value of the
selected column is reset to the value of the remaining instances.
Note: For both Add and Reset buttons, the ALTER SYSTEM command uses SCOPE=MEMORY.

EM and SPFILE Parameter Values
SCOPE=BOTH
s a
)ha
m
co
SCOPE=SPFILE
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
The SPFile
Etabbednpage
s fe displays the current values stored in your SPFILE.
e o tra tabbed page, you can add or reset parameters. However, if you select the
rtthe Current
v
As on n -
o in SPFILE mode check box, the ALTER SYSTEM command uses SCOPE=BOTH.
EApply nchanges
If this check box is not selected, SCOPE=SPFILE is used.
Click Apply to accept and generate your changes.

RAC Initialization Parameters
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfeEnables a database to be started in cluster mode. Set this to TRUE.
e o
CLUSTER_DATABASE:
rt n-tra
v Sets the number of instances in your RAC environment.
nosetting for this parameter can improve memory use.
EA proper
CLUSTER_DATABASE_INSTANCES:
CLUSTER_INTERCONNECTS: Specifies the cluster interconnect when there is more than one
interconnect. Refer to your Oracle platformspecific documentation for the use of this
parameter, its syntax, and its behavior. You typically do not need to set the
CLUSTER_INTERCONNECTS parameter. For example, do not set this parameter for the
following common configurations:
If you have only one cluster interconnect
If the default cluster interconnect meets the bandwidth requirements of your RAC
database, which is typically the case
If NIC bonding is being used for the interconnect
When OIFCFGs global configuration can specify the right cluster interconnects. It only
needs to be specified as an override for OIFCFG.
DB_NAME: If you set a value for DB_NAME in instance-specific parameter files, the setting must
be identical for all instances.
DISPATCHERS: Set this parameter to enable a shared-server configuration, that is, a server
that is configured to allow many user processes to share very few server processes.

With shared-server configurations, many user processes connect to a dispatcher. The
DISPATCHERS parameter may contain many attributes. Oracle recommends that you
configure at least the PROTOCOL and LISTENER attributes.
PROTOCOL specifies the network protocol for which the dispatcher process generates a
listening end point. LISTENER specifies an alias name for the Oracle Net Services listeners.
Set the alias to a name that is resolved through a naming method, such as a tnsnames.ora
file.
Other parameters that can affect RAC database configurations include:

ASM_PREFERRED_READ_FAILURE_GROUPS: Specifies a set of disks to be the
preferred disks from which to read mirror data copies. The values that you set for this
parameter are instance-specific and need not be the same on all instances.
GCS_SERVER_PROCESSES: This static parameter specifies the initial number of server
processes for an Oracle RAC instances Global Cache Service (GCS). The GCS
processes manage the routing of interinstance traffic among Oracle RAC instances. The
s
default number of GCS server processes is calculated based on system resources. For a
) ha
systems with one CPU, there is one GCS server process (LMSn). For systems with two
o m
y s c e
to eight CPUs, there are two GCS server processes (LMSn). For systems with more than
u n is uid
eight CPUs, the number of GCS server processes equals the number of CPUs divided
b r n t G
by 4, dropping any fractions. You can set this parameter to different values on different
instances.
n t o@ tude
me this S
INSTANCE_NAME: The instances SID. The SID identifies the instances shared memory
c i
on a host. Any alphanumeric characters can be used. The value for this parameter is
n as use
automatically set to the database unique name followed by an incrementing number
e r ton e to
during the creation of the database when using DBCA.
( e v ens
INSTANCE_NUMBER: An Oracle RAC parameter that specifies a unique number that
n n e lic
maps the instance to one free list group for each database object. This parameter must
o
e r t a b l
Ev nsfer
be set for every instance in the cluster. It is automatically defined during the creation of
r n
the database when using DBCA.
to -tra
e
REMOTE_LISTENER: This parameter resolves to a SCAN:port, resolving to a SCAN
Ev nonlistener. Oracle Database 11g release 2 (11.2) and later instances only register with
SCAN listeners as remote listeners. Upgraded databases register with SCAN listeners
as remote listeners, and also continue to register with all node listeners.
LOCAL_LISTENER: Specifies a network name that resolves to an address or address
list of Oracle Net local listeners (that is, listeners that are running on the same machine
as this instance). The address or address list is specified in the TNSNAMES.ORA file or
other address repository as configured for your system.

Parameters That Require Identical Settings
ACTIVE_INSTANCE_COUNT
ARCHIVE_LAG_TARGET
COMPATIBLE
CLUSTER_DATABASE/CLUSTER_DATABASE_INSTANCES
CONTROL_FILES
DB_BLOCK_SIZE
DB_DOMAIN
DB_FILES a
DB_NAME ha s
m )
DB_RECOVERY_FILE_DEST/DB_RECOVERY_FILE_DEST_SIZE o
c e
DB_UNIQUE_NAME y s
PARALLEL_EXECUTION_MESSAGE_SIZE run
is uid
b n t G
REMOTE_LOGIN_PASSWORD_FILE o@
n t t u de
RESULT_CACHE_MAX_SIZE
i m e is S
UNDO_MANAGEMENT as c e th n o us
r t o n t
v e s e
n (e 2012,
r t o n le
e a b
n fer
Ev nsparameters
Certain
r t oinitialization
t r a that are critical at database creation or that affect certain
ve ooperations
Edatabase - must have the same value for every instance in RAC. Specify these
n nvalues in the SPFILE, or within each init_dbname.ora file on each instance. In
parameter
the list provided in the slide, each parameter must have the same value on all instances.
Note: The setting for DML_LOCKS and RESULT_CACHE_MAX_SIZE must be identical on every
instance only if set to zero. Disabling the result cache on some instances may lead to
incorrect results.

Parameters That Require Unique Settings
Instance settings:
INSTANCE_NAME
INSTANCE_NUMBER
UNDO_TABLESPACE
CLUSTER_INTERCONNECTS
ASM_PREFERRED_READ_FAILURE_GROUPS
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
The Oracle s fethe INSTANCE_NUMBER parameter to distinguish among instances at
Eservernuses
e o tra server uses the THREAD number to assign redo log groups to specific
rt Then-Oracle
startup.
v
no To simplify administration, use the same number for both the THREAD and
Einstances.
INSTANCE_NUMBER parameters.
If you specify UNDO_TABLESPACE with Automatic Undo Management enabled, set this
parameter to a unique undo tablespace name for each instance.
Using the ASM_PREFERRED_READ_FAILURE_GROUPS initialization parameter, you can specify a
list of preferred read failure group names. The disks in those failure groups become the
preferred read disks. Thus, every node can read from its local disks. The setting for this
parameter is instance-specific, and the values need not be the same on all instances.

Quiescing RAC Databases
Use the ALTER SYSTEM QUIESCE RESTRICTED

statement from a single instance:

SQL> ALTER SYSTEM QUIESCE RESTRICTED;
You must have the Database Resource Manager feature
activated to issue the preceding statement.
The database cannot be opened by other instances after
the ALTER SYSTEM QUIESCE statement starts. s a
) ha
The ALTER SYSTEM QUIESCE RESTRICTED andoALTER m
SYSTEM UNQUIESCE statements affect all instances y s c inea
u n is uid
RAC environment. b r n t G
@ ude
Cold backups cannot be taken e nto theStdatabase
when is in a
m this
quiesced state. sci n a use

e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
To quiesce
E s fe use the ALTER SYSTEM QUIESCE RESTRICTED statement from
a RACndatabase,
onee o
rt n-tItrais not possible to open the database from any instance while the database is
instance.
v
no of being quiesced from another instance. After all the non-DBA sessions
Ein the process
become inactive, the ALTER SYSTEM QUIESCE RESTRICTED statement executes and the
database is considered to be quiesced. In a RAC environment, this statement affects all
instances.
To issue the ALTER SYSTEM QUIESCE RESTRICTED statement in a RAC environment, you
must have the Database Resource Manager feature activated, and it must have been
activated since instance startup for all instances in the cluster database. It is through the
Database Resource Manager that non-DBA sessions are prevented from becoming active.
The following conditions apply to RAC:
If you had issued the ALTER SYSTEM QUIESCE RESTRICTED statement, but the Oracle
server has not finished processing it, then you cannot open the database.
You cannot open the database if it is already in a quiesced state.
The ALTER SYSTEM QUIESCE RESTRICTED and ALTER SYSTEM UNQUIESCE statements
affect all instances in a RAC environment, not just the instance that issues the
command.
Cold backups cannot be taken when the database is in a quiesced state because the Oracle
background processes may still perform updates for internal purposes even when the
database is in a quiesced state. Also, the file headers of online data files continue to appear
as if they are being accessed. They do not look the same as if a clean shutdown were done.
Terminating Sessions on a Specific Instance
SQL> SELECT SID, SERIAL#, INST_ID

2 FROM GV$SESSION WHERE USERNAME='JMW';

SID SERIAL# INST_ID
---------- ---------- ----------
140 3340 2
SQL> ALTER SYSTEM KILL SESSION '140,3340,@2';
System altered. s a
)ha
SQL> m
co
s
ALTER SYSTEM KILL SESSION '140,3340,@2' u n isy uide
b r nt G
o@ tude
*
n t
me this S
ERROR at line 1:
c i
ORA-00031: session marked
n as for u s ekill
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Starting n EOracle
with n s fe
Database 11g Release 1, you can use the ALTER SYSTEM KILL SESSION
e o tra a session on a specific instance.
rt ton-terminate
v
statement
noillustrates this by terminating a session started on a different instance than the one
EThe slide
used to terminate the problematic session.
If the session is performing some activity that must be completed, such as waiting for a reply
from a remote database or rolling back a transaction, then Oracle Database waits for this
activity to complete, marks the session as terminated, and then returns control to you. If the
waiting lasts a minute, Oracle Database marks the session to be terminated and returns
control to you with a message that the session is marked to be terminated. The PMON
background process then marks the session as terminated when the activity is complete.
Note: You can also use the IMMEDIATE clause at the end of the ALTER SYSTEM command to
immediately terminate the session without waiting for outstanding activity to complete.

How SQL*Plus Commands Affect Instances
SQL*Plus Command Associated Instance

ARCHIVE LOG Generally affects the current instance

CONNECT Affects the default instance if no instance is specified
in the CONNECT command
HOST Affects the node running the SQL*Plus session
RECOVER Does not affect any particular instance, but rather the
database
s a
SHOW PARAMETER and Show the current instance parameter and SGA
) ha
SHOW SGA information
o m
STARTUP and Affect the current instance y s c e
SHUTDOWN u n is uid
b G
rcurrentninstance
t
SHOW INSTANCE Displays information about
n t o@ tude the
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfaffect e
Most SQL
r t o statements
t r a the current instance. You can use SQL*Plus to start and stop
ve oinnthe
instances
EUNIX-based - RAC database. You do not need to run SQL*Plus commands as root on
n systems or as Administrator on Windows-based systems. You need only the
proper database account with the privileges that you normally use for single-instance Oracle
database administration. The following are some examples of how SQL*Plus commands
affect instances:
The ALTER SYSTEM SET CHECKPOINT LOCAL statement affects only the instance to
which you are currently connected, rather than the default instance or all instances.
ALTER SYSTEM CHECKPOINT LOCAL affects the current instance.
ALTER SYSTEM CHECKPOINT or ALTER SYSTEM CHECKPOINT GLOBAL affect all
instances in the cluster database.
ALTER SYSTEM SWITCH LOGFILE affects only the current instance.
To force a global log switch, use the ALTER SYSTEM ARCHIVE LOG CURRENT
statement.
The INSTANCE option of ALTER SYSTEM ARCHIVE LOG enables you to archive each
online redo log file for a specific instance.

Transparent Data Encryption and Wallets in RAC
One wallet shared by all instances on shared storage:

No additional administration is required.

One copy of the wallet on each local storage:
Local copies need to be synchronized each time master key
is changed.
ALTER SYSTEM SET ENCRYPTION KEY 1 a

a s
m )h
Walletsc
o
Wallet Wallet
s y d e
Manual
u i
n Gu i
copy r
b Master n t
Master keys Master key
@ e key
2
Node2 e nto Stud Noden
Node1
s c im this
n aManualucopys e
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eby RAC s e
finstances
Wallets
r t oused
t r a n for Transparent Database Encryption may be a local copy of
e onwallet
Eaallvcommon - shared by multiple nodes, or a shared copy residing on shared storage that
n nodes can access.
of the
A deployment with a single wallet on a shared disk requires no additional configuration to use
Transparent Data Encryption.
If you want to use local copies, you must copy the wallet and make it available to all of the
other nodes after initial configuration. For systems using Transparent Data Encryption with
encrypted wallets, you can use any standard file transport protocol. For systems using
Transparent Data Encryption with obfuscated wallets, file transport through a secured channel
is recommended. The wallet must reside in the directory specified by the setting for the
WALLET_LOCATION or ENCRYPTION_WALLET_LOCATION parameter in sqlnet.ora. The
local copies of the wallet need not be synchronized for the duration of Transparent Data
Encryption usage until the server key is rekeyed through the ALTER SYSTEM SET KEY SQL
statement. Each time you run the ALTER SYSTEM SET KEY statement at a database
instance, you must again copy the wallet residing on that node and make it available to all of
the other nodes. To avoid unnecessary administrative overhead, reserve rekeying for
exceptional cases where you are certain that the server master key is compromised and that
not rekeying it would cause a serious security problem.

Quiz
If an instance starts in a policy-managed RAC environment and

no thread or redo log file is available, then Oracle Clusterware

automatically enables a thread of redo and allocates the redo
log files and undo if the database uses Oracle ASM or any
cluster file system and OMF is enabled.
a. True
b. False a
ha s
m )
o
c e
y s
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o a
rt n-tra
v
no is true.
E statement
The

Quiz
Which of the following statements is not true:

a. Multiple instances can open the same database

simultaneously.
b. Shutting down one instance does not interfere with other
running instances.
c. SHUTDOWN TRANSACTIONAL LOCAL will wait for other
instances transactions to finish. s a
) ha
d. Shutting down a RAC database means shutting downmall
c o
instances accessing the database. ys e is uid
r u n G
b n t
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o c
rt n-tra
v
E no c is not true.
Statement

Summary

Use Enterprise Manager Cluster Database pages

Define redo log files in a RAC environment
Define undo tablespaces in a RAC environment
Start and stop RAC databases and instances
Modify initialization parameters in a RAC environment
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Practice 4 Overview

Using operating system and password fileauthenticated

connections
Using Oracle Database authenticated connections
Stopping a complete ORACLE_HOME component stack
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Managing Backup and Recovery for RAC
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives
After completing this lesson, you should be able to configure

the following:
The RAC database to use ARCHIVELOG mode and the fast
recovery area
RMAN for the RAC environment
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

RAC and Instance Recovery
Use information
Remaster for other caches
enqueue LMS
resources Remaster recovers
1 cache GRD
resources
2
s a
Build re- )ha
SMON covery set m
co
recovers Resource s
the 3 claim
u n isy uide
database 4 b n G
rRoll forward
t
Merge failed
redo threads
n t o@ trecovery u de set
c i me this S 5
n as usetime
Recovery
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
When an n E s
instancenfailsfeand the failure is detected by another instance, the second instance
e o
rt the tra recovery steps:
performs
v n -following
no the first phase of recovery, Global Enqueue Services remasters the enqueues.
E 1. During
2. The Global Cache Services (GCS) remasters its resources. The GCS processes
remaster only those resources that lose their masters. During this time, all GCS
resource requests and write requests are temporarily suspended. However, transactions
can continue to modify data blocks as long as these transactions have already acquired
the necessary resources.
3. After enqueues are reconfigured, one of the surviving instances can grab the Instance
Recovery enqueue. Therefore, at the same time as GCS resources are remastered,
SMON determines the set of blocks that need recovery. This set is called the recovery
set. Because, with Cache Fusion, an instance ships the contents of its blocks to the
requesting instance without writing the blocks to the disk, the on-disk version of the
blocks may not contain the changes that are made by either instance. This implies that
SMON needs to merge the content of all the online redo logs of each failed instance to
determine the recovery set. This is because one failed thread might contain a hole in the
redo that needs to be applied to a particular block. So, redo threads of failed instances
cannot be applied serially. Also, redo threads of surviving instances are not needed for
recovery because SMON could use past or current images of their corresponding buffer
caches.

4. Buffer space for recovery is allocated and the resources that were identified in the
previous reading of the redo logs are claimed as recovery resources. This is done to
avoid other instances to access those resources.
5. All resources required for subsequent processing have been acquired and the Global
Resource Directory (GRD) is now unfrozen. Any data blocks that are not in recovery can
now be accessed. Note that the system is already partially available.
Then, assuming that there are past images or current images of blocks to be recovered
in other caches in the cluster database, the most recent image is the starting point of
recovery for these particular blocks. If neither the past image buffers nor the current
buffer for a data block is in any of the surviving instances caches, then SMON performs
a log merge of the failed instances. SMON recovers and writes each block identified in
step 3, releasing the recovery resources immediately after block recovery so that more
blocks become available as recovery proceeds.
After all the blocks have been recovered and the recovery resources have been released, the
system is again fully available.
s a
In summary, the recovered database or the recovered portions of the database become )h a
o
available earlier, and before the completion of the entire recovery sequence.cThism makes the

system available sooner and it makes recovery more scalable.
n i sys uide
Note: The performance overhead of a log merge is proportionalrto
b uthe number
t G of failed
redo logs n
@ tude each instance.
instances and to the size of the amount of redo written in the for
t o
i m en is S
a s c e th
o n n o us
e r t e t
v
(e licen s
n n
e r t o
a b le
n Ev nsfer
e r to -tra
Ev non

Instance Recovery and Database Availability
Full A G H
Partial B F
s a
4 )ha
2 m
co
1 3 s
None C D E
u n isy uide
r nt G
Elapsed time @b de
n t o t u
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
The graphic
Eillustrates
n s fethe degree of database availability during each step of Oracle instance
e o
rt n-tra
recovery:
v
no Application Clusters is running on multiple nodes.
E A. Real
B. Node failure is detected.
C. The enqueue part of the GRD is reconfigured; resource management is redistributed to
the surviving nodes. This operation occurs relatively quickly.
D. The cache part of the GRD is reconfigured and SMON reads the redo log of the failed
instance to identify the database blocks that it needs to recover.
E. SMON issues the GRD requests to obtain all the database blocks it needs for recovery.
After the requests are complete, all other blocks are accessible.
F. The Oracle server performs roll forward recovery. Redo logs of the failed threads are
applied to the database, and blocks are available right after their recovery is completed.
G. The Oracle server performs rollback recovery. Undo blocks are applied to the database
for all uncommitted transactions.
H. Instance recovery is complete and all data is accessible.
Note: The dashed line represents the blocks identified in step 2 of the previous slide. Also,
the dotted steps represent the ones identified in the previous slide.

Instance Recovery and RAC
Instance startup
Instance + Instance
crashes crash recovery opens
FAST_START_MTTR_TARGET
Instance
starts
Rolling
Instance forward a
ends a s
Instance recovery
m )h
crashes first pass + lock claim
s co

FAST_START_MTTR_TARGET
u n isy uide
b r nt G
Instance
recovery n t o@ tude
me this S
V$INSTANCE_RECOVERY.ESTD_CLUSTER_AVAILABLE_TIME
starts c i
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe
E nsenvironment,
r t o
In a single-instance
t r a the instance startup combined with the crash recovery time
e on-by the setting of the FAST_START_MTTR_TARGET initialization parameter. You can
isvcontrolled
Eset n if you want incremental checkpointing to be more aggressive than the autotuned
its value
checkpointing. However, this is at the expense of a much higher I/O overhead.
In a RAC environment, including the startup time of the instance in this calculation is useless
because one of the surviving instances is doing the recovery.
In a RAC environment, it is possible to monitor the estimated target, in seconds, for the
duration from the start of instance recovery to the time when GCD is open for lock requests
for blocks not needed for recovery. This estimation is published in the V$INSTANCE_RECOVERY
view through the ESTD_CLUSTER_AVAILABLE_TIME column. Basically, you can monitor the
time your cluster is frozen during instance-recovery situations.
In a RAC environment, the FAST_START_MTTR_TARGET initialization parameter is used to
bound the entire instance-recovery time, assuming it is instance recovery for single-instance
death.
Note: If you really want to have short instance recovery times by setting
FAST_START_MTTR_TARGET, you can safely ignore the alert log messages advising you to
raise its value.

Instance Recovery and RAC
Use parallel instance recovery.

Set PARALLEL_MIN_SERVERS.
Use asynchronous input/output (I/O).
Increase the size of the default buffer cache.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe you can use to make sure that instance recovery in your RAC
E guidelines
e o
Here are some a
rt n-istrfaster:
v
environment
E Use noparallel instance recovery by setting RECOVERY_PARALLISM.
Set PARALLEL_MIN_SERVERS to CPU_COUNT-1. This will prespawn recovery slaves at
startup time.
If a system fails when there are uncommitted parallel DML or DDL transactions, you can
speed up transaction recovery during startup by setting the
FAST_START_PARALLEL_ROLLBACK parameter.
Using asynchronous I/O is one of the most crucial factors in recovery time. The first-
pass log read uses asynchronous I/O.
Instance recovery uses 50 percent of the default buffer cache for recovery buffers. If this
is not enough, some of the steps of instance recovery will be done in several passes.
You should be able to identify such situations by looking at your alert.log file. In that
case, you should increase the size of your default buffer cache.

Protecting Against Media Failure
Archived Archived
log files log files
Database s a
Mirrored backups )ha
disks m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E provides
n s feyou with methods to avoid or to reduce down time due to a failure of
e o
Although RAC
tranot all) of your instances, you must still protect the database itself, which is
rt moren-(but
v
one or
Esharednbyoall the instances. This means that you need to consider disk backup and recovery
strategies for your cluster database just as you would for a nonclustered database.
To minimize the potential loss of data due to disk failures, you may want to use disk mirroring
technology (available from your server or disk vendor). As in nonclustered databases, you can
have more than one mirror if your vendor allows it, to help reduce the potential for data loss
and to provide you with alternative backup strategies. For example, with your database in
ARCHIVELOG mode and with three copies of your disks, you can remove one mirror copy and
perform your backup from it while the two remaining mirror copies continue to protect ongoing
disk activity. To do this correctly, you must first put the tablespaces into backup mode and
then, if required by your cluster or disk vendor, temporarily halt disk operations by issuing the
ALTER SYSTEM SUSPEND command. After the statement completes, you can break the mirror
and then resume normal operations by executing the ALTER SYSTEM RESUME command and
taking the tablespaces out of backup mode.

Media Recovery in Oracle RAC
Media recovery must be user-initiated through a client

application.
In these situations, use RMAN to restore backups of the
data files and then recover the database.
RMAN media recovery procedures for RAC do not differ
substantially from those for single-instance environments.
The node that performs the recovery must be able to s a
restore all of the required data files. ) h a
c e
That node must also be able to either read all required
om
y s
nis backups.
archived redo logs on disk or restore themufrom uid r nt G
b
When recovering a database with encrypted
n t o@ tudetablespaces,
the Oracle Wallet must be opened
c i me tafter h is Sdatabase mount
and before you open the n s se
adatabase.
o n o u
r t t
( e ve ense
o n n e2012,
Copyright
v e rt rab
Mediato n E must
recovery n s fbee user-initiated through a client application, whereas instance recovery
e r n-traperformed by the database. In these situations, use RMAN to restore backups
isvautomatically
no files and then recover the database. The procedures for RMAN media recovery in
Eof the data
Oracle RAC environments do not differ substantially from the media recovery procedures for
single-instance environments.
The node that performs the recovery must be able to restore all of the required data files. That
node must also be able to either read all the required archived redo logs on disk or be able to
restore them from backups. Each instance generates its own archive logs that are copies of
its dedicated redo log group threads. It is recommended that Automatic Storage Management
(ASM) or a cluster file system be used to consolidate these files.
When recovering a database with encrypted tablespaces (for example, after a SHUTDOWN
ABORT or a catastrophic error that brings down the database instance), you must open the
Oracle Wallet after database mount and before you open the database, so the recovery
process can decrypt data blocks and redo.

Parallel Recovery in RAC
Oracle Database automatically selects the optimum degree

of parallelism for:
Instance recovery
Crash recovery
Archived redo logs are applied using an optimal number of
parallel processes based on the availability of CPUs.
With RMANs RESTORE and RECOVER commands, the s a
following three stages of recovery can use parallelism: ) ha
Restoring data files c om
i s ys ide
Applying incremental backups
r u n Gu
Applying archived redo logs @ b ent
e nto Stud
To disable parallel instancecand
s im crash
e t his recovery, set the
RECOVERY_PARALLELISM
n na parameter
u s to 0.
rto se t o
n (e 2012,
r t o n le
e a b
n s f er
Ev automatically
Oracle
r t o
Database
t r a n selects the optimum degree of parallelism for instance and
ve recovery.
Ecrash on- Oracle Database applies archived redo logs using an optimal number of
parallelnprocesses based on the availability of CPUs. With RMANs RESTORE and RECOVER
commands, Oracle Database automatically uses parallelism for the following three stages of
recovery:
Restoring Data Files: When restoring data files, the number of channels you allocate in
the RMAN recover script effectively sets the parallelism that RMAN uses. For example,
if you allocate five channels, you can have up to five parallel streams restoring data files.
Applying Incremental Backups: Similarly, when you are applying incremental
backups, the number of channels you allocate determines the potential parallelism.
Applying Archived Redo Logs with RMAN: Oracle Database automatically selects
the optimum degree of parallelism based on available CPU resources.
To disable parallel instance and crash recovery on a system with multiple CPUs, set the
RECOVERY_PARALLELISM parameter to 0.
Use the NOPARALLEL clause of the RMAN RECOVER command or ALTER DATABASE
RECOVER statement to force the RAC database to use nonparallel media recovery.

Archived Log File Configurations
s a
)ha
m
co
s
u n isy uide
Shared storage scheme: Local archive b NFS
rwith n t G
Archive logs from each scheme:
n t o@ u de can
Each tinstance
instance are written to the read
m e mounted i s Sarchive
c i h
t of all instances.
same file location.
n as destinations
u s e
r t o n t o
( e ve ense
o n n e2012,
Copyright
v e rt rab
n s fe operations involving archived log files, the Oracle server
E andnrecovery
During
e obackup
rt nthe a destinations and names from the control file. If you use RMAN, the
trfile
v
determines -
Earchivednolog file path names can also be stored in the optional recovery catalog. However, the
archived log file path names do not include the node name, so RMAN expects to find the files
it needs on the nodes where the channels are allocated.
If you use a supported shared storage scheme, your instances can all write to the same
archive log destination. Backup and recovery of the archive logs are easy because all logs
are located in the same directory.
If a shared storage location is not available, Oracle recommends that local archive log
destinations be created for each instance with NFS-read mount points to all other instances.
This is known as the local archive with network file system (NFS) scheme. During backup,
you can either back up the archive logs from each host or select one host to perform the
backup for all archive logs. During recovery, one instance may access the logs from any host
without having to first copy them to the local destination. The LOG_ARCHIVE_FORMAT
parameter supports the %t variable that embeds the unique thread number into the name of
the archive logs so that each node generates unique names.
Using either scheme, you may want to provide a second archive destination to avoid single
points of failure.

RAC and the Fast Recovery Area
Fast
recovery
area s a
) ha
o m
y s c e
u
Certifiedn isNFS uid
Cluster file system b
rdirectory
n t G
n t o@ tude
c
ASM i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Erecovery
n s fearea in RAC, you must place it on an ASM disk group, a cluster file
To use
e oa fast
rt or non-tarashared directory that is configured through certified NFS for each RAC
v
system,
noThat is, the fast recovery area must be shared among all the instances of a RAC
Einstance.
database. In addition, set the DB_RECOVERY_FILE_DEST parameter to the same value on all
instances.
Oracle Enterprise Manager enables you to set up a fast recovery area. To use this feature:
1. From the Cluster Database Home page, click the Maintenance tab.
2. Under the Backup/Recovery options list, click Configure Recovery Settings.
3. Specify your requirements in the Flash Recovery Area section of the page.
Note: For Oracle Database 11g Release 2 (11.2), the flash recovery area has been renamed
fast recovery area. Oracle Enterprise Manager, however, still uses the older vocabulary on
its webpages.

RAC Backup and Recovery Using EM
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E the n s fe Database backup and recoveryrelated tasks by clicking the
You can
e o access
rt ntab Cluster
trona the Cluster Database Home page. On the Availability tabbed page, you can
v
Availability -
Eperformnaorange of backup and recovery operations using RMAN, such as scheduling
backups, performing recovery when necessary, and configuring backup and recovery
settings. Also, there are links related to Oracle Secure Backup and Service management.

Configuring RAC Recovery Settings with EM
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EEnterprise
n s feManager to configure important recovery settings for your cluster
You can
e o
rt Onuse
ra Database Home page, click the Availability tab, and then click the Recovery
tthe
v
database. n -
ESettingsnolink. From here, you can ensure that your database is in ARCHIVELOG mode and
configure flash recovery settings.
With a RAC database, if the Archive Log Destination setting is not the same for all instances,
the field appears blank, with a message indicating that instances have different settings for
this field. In this case, entering a location in this field sets the archive log location for all
instances of the database. You can assign instance-specific values for an archive log
destination by using the Initialization Parameters page.
Note: You can run the ALTER DATABASE SQL statement to change the archiving mode in
RAC as long as the database is mounted by the local instance but not open in any instances.
You do not need to modify parameter settings to run this statement. Set the initialization
parameters DB_RECOVERY_FILE_DEST and DB_RECOVERY_FILE_DEST_SIZE to the same
values on all instances to configure a fast recovery area in a RAC environment.

Archived Redo File Conventions in RAC
Variable Description Example

%t Thread number, not padded log_1
%T Thread number, left-zero-padded log_0001
%s Log sequence number, not padded log_251
s a
%S Log sequence number, left-zero-padded log_0000000251
)ha
m
co
s
%r Resetlogs identifier log_23452345
u n isy uide
b r nt G
%R Padded resetlogs identifier
n t o@ tudelog_0023452345
i m e is S
%t_%s_%r Using multiple variables s
a c e th log_1_251_23452345
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n Ev redo s f er
For any
r t oarchived
t r a n log configuration, uniquely identify the archived redo logs with the
ve on-
ELOG_ARCHIVE_FORMAT parameter. The format of this parameter is operating system specific
n include text strings, one or more variables, and a file name extension.
and it can
All of the thread parameters, in either uppercase or lowercase, are mandatory for RAC. This
enables the Oracle database to create unique names for archive logs across the incarnation.
This requirement is in effect when the COMPATIBLE parameter is set to 10.0 or greater. Use
the %R or %r parameter to include the resetlogs identifier to avoid overwriting the logs from a
previous incarnation. If you do not specify a log format, the default is operating system
specific and includes %t, %s, and %r.
As an example, if the instance associated with redo thread number 1 sets
LOG_ARCHIVE_FORMAT to log_%t_%s_%r.arc, then its archived redo log files are named
as:
log_1_1000_23435343.arc
log_1_1001_23452345.arc
log_1_1002_23452345.arc
...
Note: The LOG_ARCHIVE_FORMAT parameter will have no effect if Oracle Managed Files
(OMF) has been implemented for the Oracle RAC database.

Configuring RAC Backup Settings with EM
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsettings
s fe can be configured using Enterprise Manager. On the Database
e o
Persistent backup
rt Home ra click the Availability tab, and then click the Backup Settings link. You can
tpage,
v
Control n -
nodisk settings, such as the directory location of your disk backups and level of
Econfigure
parallelism. You can also choose the default backup type:
Backup set
Compressed backup set
Image copy
You can also specify important tape-related settings, such as the number of available tape
drives and vendor-specific media management parameters.

Oracle Recovery Manager
RMAN provides the following

Recovery benefits for Real Application

Manager Clusters:
Recovery Can read cluster files or
catalog
Archived ASM files with no
Oracle
log files configuration changes
Server
process Can access multiple a
Stored
archive log destinationsha s
scripts )
c om
i s ys ide
Oracle r u n Gu
database @ b ent
Backup
e n to Stud
storage Snapshot
s c im this
control file
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E provides
n s fe RMAN for backing up and restoring the database. RMAN enables
Oracle
e oDatabase
trarestore, and recover data files, control files, SPFILEs, and archived redo logs.
rt backnup,
v
you to -
E no
You can run RMAN from the command line or you can use it from the Backup Manager in
Enterprise Manager. In addition, RMAN is the recommended backup and recovery tool if you
are using ASM. RMAN can use stored scripts, interactive scripts, or an interactive GUI front
end. When using RMAN with your RAC database, use stored scripts to initiate the backup and
recovery processes from the most appropriate node.
If you use different Oracle Home locations for your RAC instances on each of your nodes,
create a snapshot control file in a location that exists on all your nodes. The snapshot control
file is needed only on the nodes on which RMAN performs backups.
You can use either a cluster file or a local directory that exists on each node in your cluster.
Here is an example:
RMAN> CONFIGURE SNAPSHOT CONTROLFILE TO
'/oracle/db_files/snaps/snap_prod1.cf';
For recovery, you must ensure that each recovery node can access the archive log files from
all instances by using one of the archive schemes discussed earlier, or make the archived
logs available to the recovering instance by copying them from another location.

Configuring RMAN
Snapshot Control File Location
The snapshot control file path must be valid on every node
from which you might initiate an RMAN backup.

Configure the snapshot control file location in RMAN.
Determine the current location:
RMAN> SHOW SNAPSHOT CONTROLFILE NAME;
/u01/app/oracle/product/11.2.0/dbhome_1/dbs/snap_prod.f
You can use ASM or a shared file system location if you s a

prefer: ) ha
m
co
RMAN> CONFIGURE SNAPSHOT CONTROLFILE NAME TO s
'+FRA/SNAP/snap_prod.cf';
u n isy uide
b r TO nt G
o@ tude
RMAN> CONFIGURE SNAPSHOT CONTROLFILE NAME
'/ocfs2/oradata/dbs/scf/snap_prod.cf'; n t
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E control s e
ffile
r t o
The snapshot
t r a n is a copy of a database control file created in an operating system
ve location
Especific n- by RMAN. RMAN creates the snapshot control file so that it has a consistent
versionnofoa control file to use when either resynchronizing the recovery catalog or backing up
the control file. You can also create a snapshot control file by entering the following at the
RMAN prompt: DUPLICATE FROM ACTIVE. You can specify a cluster file system or a raw
device destination for the location of your snapshot control file. This file is shared across all
nodes in the cluster and must be accessible by all nodes in the cluster.
You can change the configured location of the snapshot control file. For example, on Linux
and UNIX systems you can change the snapshot control file location using the CONFIGURE
SNAPSHOT CONTROLFILE NAME RMAN command. This command sets the configuration for
the location of the snapshot control file for every instance of your cluster database. Therefore,
ensure that the location specified exists on all nodes that perform backups. The CONFIGURE
command creates persistent settings across RMAN sessions. Therefore, you do not need to
run this command again unless you want to change the location of the snapshot control file.
To delete a snapshot control file you must first change the snapshot control file location, then
delete the file at the older location, as follows:
CONFIGURE SNAPSHOT CONTROLFILE NAME TO 'new_name';
DELETE COPY OF CONTROLFILE;

Configuring Control File and SPFILE Autobackup
RMAN automatically creates a control file and SPFILE

backup after the BACKUP or COPY command:

RMAN> CONFIGURE CONTROLFILE AUTOBACKUP ON;
Change default location:

RMAN> CONFIGURE CONTROLFILE AUTOBACKUP FORMAT FOR DEVICE
TYPE DISK TO '+DATA';
s a
Location must be available to all nodes in your RAC ) h
a
database. c om
i s ys ide
r u n Gu
@ b ent
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
If you set n e
E nsfCONTROLFILE AUTOBACKUP to ON, RMAN automatically creates a control
r t o a
CONFIGURE
t r
veand anonSPFILE
Efile - backup after you run the BACKUP or COPY command. RMAN can also
n restore an SPFILE if this is required to start an instance to perform recovery.
automatically
This means that the default location for the SPFILE must be available to all nodes in your
RAC database.
These features are important in disaster recovery because RMAN can restore the control file
even without a recovery catalog. RMAN can restore an autobackup of the control file even
after the loss of both the recovery catalog and the current control file.
You can change the default location that RMAN gives to this file with the CONFIGURE
CONTROLFILE AUTOBACKUP FORMAT command. If you specify an absolute path name in this
command, this path must exist identically on all nodes that participate in backups.
Note: RMAN performs CONTROL FILE AUTOBACKUP on the first allocated channel. When
you allocate multiple channels with different parameters (especially if you allocate a channel
with the CONNECT command), you must determine which channel will perform the automatic
backup. Always allocate the channel for the connected node first.

Crosschecking on Multiple RAC Clusters Nodes
When crosschecking on multiple nodes, make sure that all

backups can be accessed by every node in the cluster.

This allows you to allocate channels at any node in the
cluster during restore or crosscheck operations.
Otherwise, you must allocate channels on multiple nodes
by providing the CONNECT option to the CONFIGURE
CHANNEL command. a
a s
If backups are not accessible because no channel was) h
configured on the node that can access those backups, c om
s y s de
then those backups are marked EXPIRED.un i u i

br ent G
@
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfone multiple RAC nodes, configure the cluster so that all backups can be
e o
Whentcrosschecking
r by tra node, regardless of which node created the backup. When the cluster is
v
accessed n -every
no this way, you can allocate channels at any node in the cluster during restore or
Econfigured
crosscheck operations.
If you cannot configure the cluster so that each node can access all backups, then during
restore and crosscheck operations, you must allocate channels on multiple nodes by
providing the CONNECT option to the CONFIGURE CHANNEL command, so that every backup
can be accessed by at least one node. If some backups are not accessible during crosscheck
because no channel was configured on the node that can access those backups, then those
backups are marked EXPIRED in the RMAN repository after the crosscheck.
For example, you can use CONFIGURE CHANNEL ... CONNECT in an Oracle RAC
configuration in which tape backups are created on various nodes in the cluster and each
backup is accessible only on the node on which it is created.

Channel Connections to Cluster Instances
When backing up, each allocated channel can connect to a
different instance in the cluster.
Instances to which the channels connect must be either all

mounted or all open.
When choosing a channel to use, RMAN gives preference
to the nodes with faster access to the data files that you
want to back up.
CONFIGURE DEFAULT DEVICE TYPE TO sbt;
s a
CONFIGURE DEVICE TYPE sbt PARALLELISM 3;
)ha
CONFIGURE CHANNEL 1 DEVICE TYPE sbt CONNECT='sys/rac@orcl1'; m
co
s
isy uide
CONFIGURE CHANNEL 2 DEVICE TYPE sbt CONNECT='sys/rac@orcl2';
CONFIGURE CHANNEL 3 DEVICE TYPE sbt CONNECT='sys/rac@orcl3';
u n
r nt G
OR b
n t o@ tude
c i me t3; h is S
CONFIGURE DEVICE TYPE sbt PARALLELISM
n assbtuCONNECT='sys/rac@bkp_serv';
s e
ton e to
CONFIGURE CHANNEL DEVICE TYPE
e r
v ens
( e
o n n e2012,
Copyright
v e rt rab
n E backups
n s fein parallel, RMAN channels can connect to a different instance in the
e o
Whentmaking
tra in the slide illustrate two possible configurations:
r Thenexamples
v
cluster.
E Ifnyou -
o want to dedicate channels to specific instances, you can control at which instance
the channels are allocated by using separate connect strings for each channel
configuration as shown by the first example.
If you define a special service for your backup and recovery jobs, you can use the
second example shown in the slide. If you configure this service with load balancing
turned on, then the channels are allocated at a node as decided by the load balancing
algorithm.
During a backup, the instances to which the channels connect must be either all mounted or
all open. For example, if the orcl1 instance has the database mounted whereas the orcl2
and orcl3 instances have the database open, then the backup fails.
In some RAC database configurations, some cluster nodes have faster access to certain data
files than to other data files. RMAN automatically detects this, which is known as node affinity
awareness. When deciding which channel to use to back up a particular data file, RMAN
gives preference to the nodes with faster access to the data files that you want to back up.

RMAN Channel Support for the Grid
RAC allows the use of nondeterministic connect strings.

RMAN can use connect strings that are not bound to a

specific instance in the Grid environment.
It simplifies the use of parallelism with RMAN in a RAC
environment.
It uses the load-balancing characteristics of the Grid
environment. s a
a
Channels connect to RAC instances that are the least m) h
loaded. s co
u n isy uide
CONFIGURE DEFAULT DEVICE TYPE TO sbt;@b
r nt G
n t o 3; tude
me this S
CONFIGURE DEVICE TYPE sbt PARALLELISM
c i
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Ethe use n s fenondeterministic connect strings that can connect to different instances
e o
RAC allows
rt on RAC a of
trfeatures,
v
based
o n -
Epollingnmechanism such as load balancing. Therefore, to support RAC, the RMAN
no longer depends on deterministic connect strings, and makes it possible
to use RMAN with connect strings that are not bound to a specific instance in the Grid
environment. Previously, if you wanted to use RMAN parallelism and spread a job between
many instances, you had to manually allocate an RMAN channel for each instance. To use
dynamic channel allocation, you do not need separate CONFIGURE CHANNEL CONNECT
statements anymore. You only need to define your degree of parallelism by using a command
such as CONFIGURE DEVICE TYPE disk PARALLELISM, and then run backup or restore
commands. RMAN then automatically connects to different instances and does the job in
parallel. The Grid environment selects the instances that RMAN connects to, based on load
balancing. As a result of this, configuring RMAN parallelism in a RAC environment becomes
as simple as setting it up in a non-RAC environment. By configuring parallelism when backing
up or recovering a RAC database, RMAN channels are dynamically allocated across all RAC
instances.
Note: RMAN has no control over the selection of the instances. If you require a guaranteed
connection to an instance, you should provide a connect string that can connect only to the
required instance.

RMAN Default Autolocation
Recovery Manager autolocates the following files:

Backup pieces
Archived redo logs during backup
Data file or control file copies
If local archiving is used, a node can read only those
archived logs that were generated on that node.
When restoring, a channel connected to a specific node s a
h a
m)
restores only those files that were backed up to the node.
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe
E nsautomatically
Recovery
r t o Manager
t r a discovers which nodes of a RAC configuration can access
vefiles thatn-you want to back up or restore. Recovery Manager autolocates the following
Ethe
files: no
Backup pieces during backup or restore
Archived redo logs during backup
Data file or control file copies during backup or restore
If you use a noncluster file system local archiving scheme, a node can read only those
archived redo logs that were generated by an instance on that node. RMAN never attempts to
back up archived redo logs on a channel that it cannot read.
During a restore operation, RMAN automatically performs the autolocation of backups. A
channel connected to a specific node attempts to restore only those files that were backed up
to the node. For example, assume that log sequence 1001 is backed up to the drive attached
to node 1, whereas log 1002 is backed up to the drive attached to node 2. If you then allocate
channels that connect to each node, the channel connected to node 1 can restore log 1001
(but not 1002), and the channel connected to node 2 can restore log 1002 (but not 1001).

Distribution of Backups
Several possible backup configurations for RAC:

A dedicated backup server performs and manages

backups for the cluster and the cluster database.
One node has access to a local backup appliance and
performs and manages backups for the cluster database.
Each node has access to a local backup appliance and
can write to its own local backup media. s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n e
E nthesfbackup
r o
Whentconfiguring
t r a options for RAC, you have several possible configurations:
Eve Network
n o n- backup server: A dedicated backup server performs and manages backups
for the cluster and the cluster database. None of the nodes have local backup
appliances.
One local drive: One node has access to a local backup appliance and performs and
manages backups for the cluster database. All nodes of the cluster should be on a
cluster file system to be able to read all data files, archived redo logs, and SPFILEs. It is
recommended that you do not use the noncluster file system archiving scheme if you
have backup media on only one local drive.
Multiple drives: Each node has access to a local backup appliance and can write to its
own local backup media.
In the cluster file system scheme, any node can access all the data files, archived redo logs,
and SPFILEs. In the noncluster file system scheme, you must write the backup script so that
the backup is distributed to the correct drive and path for each node. For example, node 1 can
back up the archived redo logs whose path names begin with /arc_dest_1, node 2 can
back up the archived redo logs whose path names begin with /arc_dest_2, and node 3 can
back up the archived redo logs whose path names begin with /arc_dest_3.

Shared Storage Backup Scheme: One Local Drive
RMAN> CONFIGURE DEVICE TYPE sbt PARALLELISM 1;

RMAN> CONFIGURE DEFAULT DEVICE TYPE TO sbt;
RMAN> BACKUP DATABASE PLUS ARCHIVELOG DELETE INPUT;
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe scheme, each node in the cluster has read access to all the data
Estoragensbackup
e t o
In a shared
rarchived ra logs, and SPFILEs. This includes Automated Storage Management
tredo
v
files, n -
o file systems, and Network Attached Storage (NAS).
E(ASM),ncluster
When backing up to only one local drive in the cluster file system backup scheme, it is
assumed that only one node in the cluster has a local backup appliance such as a tape drive.
In this case, run the following one-time configuration commands:
Because any node performing the backup has read/write access to the archived redo logs
written by the other nodes, the backup script for any node is simple:
BACKUP DATABASE PLUS ARCHIVELOG DELETE INPUT;
In this case, the tape drive receives all data files, archived redo logs, and SPFILEs.

Shared Storage Backup Scheme: Multiple Drives


CONFIGURE CHANNEL 1 DEVICE TYPE sbt CONNECT 'usr1/pwd1@n1';
CONFIGURE CHANNEL 2 DEVICE TYPE sbt CONNECT 'usr2/pwd2@n2';
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe drives in the shared storage backup scheme, it is assumed that
E up ntosmultiple
e o
Whentbacking
tracluster has its own local tape drive. Perform the following one-time
rnodenin-the
v
each
no so that one channel is configured for each node in the cluster. For example,
Econfiguration
enter the following at the RMAN prompt:
CONFIGURE CHANNEL 1 DEVICE TYPE sbt CONNECT 'user1/passwd1@node1';
CONFIGURE CHANNEL 2 DEVICE TYPE sbt CONNECT 'user2/passwd2@node2';
Similarly, you can perform this configuration for a device type of DISK. The following backup
script, which you can run from any node in the cluster, distributes the data files, archived redo
logs, and SPFILE backups among the backup drives:

Restoring and Recovering
Media recovery may require one or more archived log files

from each thread.

The RMAN RECOVER command automatically restores and
applies the required archived logs.
Archive logs may be restored to any node performing the
restore and recover operation.
Logs must be readable from the node performing the s a
restore and recovery activity. ) ha
o m
Recovery processes request additional threadsys c e
enabled
during the recovery period. u n is uid
b r n t G
o@ tuno
Recovery processes notify you of tthreads
n delonger
S
ime his
needed because they were disabled.
s c et
a
n o us
r t o n t
v e s e
n (e 2012,
r t o n le
e a b
Mediato n
recovery
fer that is accessed by RAC may require at least one archived log
Ev ofnasdatabase
v
fileefor - tra However, if a threads online redo log contains enough recovery
r eachnthread.
no restoring archived log files for any thread is unnecessary.
Einformation,
If you use RMAN for media recovery and you share archive log directories, you can change
the destination of the automatic restoration of archive logs with the SET clause to restore the
files to a local directory of the node where you begin recovery. If you backed up the archive
logs from each node without using a central media management system, you must first
restore all the log files from the remote nodes and move them to the host from which you will
start recovery with RMAN. However, if you backed up each nodes log files using a central
media management system, you can use RMANs AUTOLOCATE feature. This enables you to
recover a database by using the local tape drive on the remote node.
If recovery reaches a time when an additional thread was enabled, the recovery process
requests the archived log file for that thread. If you are using a backup control file, when all
archive log files are exhausted, you may need to redirect the recovery process to the online
redo log files to complete recovery. If recovery reaches a time when a thread was disabled,
the process informs you that the log file for that thread is no longer needed.

Quiz
Which of the following statements regarding media recovery in

RAC is not true?

a. Media recovery must be user-initiated through a client
application.
b. RMAN media recovery procedures for RAC are quite
different from those for single-instance environments.
c. The node that performs the recovery must be able to s a
restore all the required data files. ) ha
o m
d. The recovering node must be able to either read y s c e
all
required archived redo logs on disk or restore
u n isthemufrom
id

br ent G
backups. o @ d
e nt S tu
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o b
rt n-tra
v
E no b is not correct.
Statement

Quiz
To use a fast recovery area in RAC, you must place it on an

ASM disk group, a cluster file system, or on a shared directory

that is configured through certified NFS for each RAC instance.
a. True
b. False
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o a
rt n-tra
v
no above is true.
E statement
The

Summary
In this lesson, you should have learned how to configure the

following:
The RAC database to use ARCHIVELOG mode and the fast
recovery area
RMAN for the RAC environment
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Practice 5 Overview

Configuring the archive log mode

Configuring specific instance connection strings
Configuring RMAN and performing parallel backups
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non
Global Resource Management Concepts
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives
After completing this lesson, you should be able to describe:

The need for global concurrency control

Global Resource Directory
How global resources are managed
RAC global resource access coordination
Global enqueue and instance lock management
Global buffer cache management s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Need for Global Concurrency Control
Oracle requires concurrency control because it is a multi-

user system.
Single-instance Oracle provides concurrency control:
Latches or mutexes for memory structures
Enqueues for resource control
Buffer cache pins for cache management
In RAC, structures and resources may be accessed by or s a
modified by a session running on any database instance. ) ha
c om
ys ide
RAC, therefore, requires additional global concurrency
i s
run t Gu
controls to mediate access across instances.
b n
Global locks control library and row
n t o@ de
cacheuaccess.
t
Global enqueues control resource
c i is S
me thaccess.
Cache fusion controls n as ucache
buffer s e access.
n r to e to
e e
v ens
(
n e2012,
Copyright licOracle and/or its affiliates. All rights reserved.
o n
rt rab l
v e
Emutexes e
n n s fmay
Latches
r t o and
t r a only protect access to memory structures if they are accessed by
ve onin-the same instance.
Eprocesses
In RAC,nlatches and mutexes are still used, but for global concurrency control, some
additional global enqueues are used to provide protection across instances.

Global Resource Directory (GRD)
An object under global concurrency control is called a

resource.
Resource metadata is held in the Global Resource
Directory (GRD).
Global enqueue resources are used for enqueues and locks.
Global cache resources are used for buffer cache control.
The GRD is distributed among all active instances of each s a
database or ASM environment. ) ha
Each currently managed GRD resource has: ysc e
om
n is id
A master metadata structure r
b ent u Gu
@
One or more shadow metadata structures
to ud t
e n S
The GRD uses memory from
s c m shared
ithe
e t his pool.
n na us
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsstructure fe contains information about the state of the related resource for
A master
e t o metadata
rinstance tinrawhich that resource resides. A shadow metadata structure only contains
v
each n -
no about the state of the related resource in the instance containing the shadow
Einformation
metadata.

Global Resource Management
After first access of a globally managed entity by any

instance, a global resource is allocated.

An internal algorithm is used to decide which instance
should contain the master metadata structure for that
entity.
This instance is known as the resource master.
The resource mastering instance may be any active instance a
of the database or ASM environment. ha s
) m
Subsequent access to an entity from another instance s co for
which resource master metadata exists causes i s y i d e
r u n resource
G u
shadow metadata to be allocated in @thebrequesting
e n t
o
t theSmaster
tu d
ento
instance and updates to be done
im this
metadata.
s
a usec
n
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E mastering
n s fe instance is the instance containing the master metadata used to
e o
The resource
tra control for a specific entity.
rt concurrency
v
manage n -
no will be the resource master for some of the database entities.
EEach instance
The resource shadowing instance is any instance containing shadow metadata used to
manage concurrency control for a specific entity. Each instance will contain shadow
resources for entities it has accessed and for which it is not the resource master.

Global Resource Remastering
Remastering is the process of allocating control of the

master metadata for a specific entity to another instance.

Instance-level or lazy remastering occurs when:
A new instance of the same database or ASM starts
A current instance is shut down gracefully
File affinity remastering occurs when:
Requests to access blocks in a data file occur frequently s a
from an instance, and the resource masters for the blocks ) ha
o m
are often held by other instances sc
Object-affinity remastering occurs when: unis
y
u ide
b r n t G
Requests to access blocks in a data
t o@ object d
u e frequently
occur
from an instance, and the resource n
e masters t
S for the blocks
c i m h i s
are often held by otherainstances
n s se t
o n o u
r t t
( e ve ense
o n n e2012,
Copyright
v e rt rab
n Einstance s e
fstarts,
Whenta
r o new
t r a n remastering is not done immediately. Instead it is done
ve obased
Egradually, n- on which instances are accessing which resources (hence the term lazy).
n
When an instance shuts down gracefullymeaning normal, immediate, or transactional
then resources mastered by the terminating instance are handed off to the surviving instances
by using an optimized internal algorithm designed to minimize the remastering and
subsequent concurrency control overheads.
The decision to perform file-affinity or object-affinity remastering is made automatically when
an internal threshold is reached.

Global Resource Recovery
When one or more but not all instances fail:

The failing instance(s) resource masters are lost

Any resource master that had a shadow in a surviving
instance must be recovered
The surviving instances can rebuild resource master
metadata for a specific resource, by aggregating details
from surviving shadow metadata for the same resource. a
ha s
Global locks and enqueue metadata are done first, )
o m
followed by global cache metadata.
y s c e
n
The rebuilding results in each surviving instance
u is mastering
u id
r t G
some of the recovered resource master@bmetadata.
d e n
e nto Stu
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E done n s fe because Oracle must know who has access to which resource in
Enqueues
e o are
rtfor recovery first,
tra to proceed. A look into the RAC database alert log shows global resource
v
order n -
o instance recovery:
Eactivitynduring
...
lmon registered with NM - instance number 1 (internal mem no 0)
Reconfiguration started (old inc 0, new inc 2)
List of instances:
1 (myinst: 1)
Global Resource Directory frozen
Communication channels reestablished
Master broadcasted resource hash value bitmaps
Non-local Process blocks cleaned out
Set master node info
Submitted all remote-enqueue requests
Dwn-cvts replayed, VALBLKs dubious
All grantable enqueues granted
Post SMON to start 1st pass IR
Submitted all GCS remote-cache requests
Post SMON to start 1st pass IR
Reconfiguration complete
...

Global Resource Background Processes
ACMS: Atomic Control file to Memory Service

LMHB: Monitors LMON, LMD, and LMSn processes

LMD0: Requests global enqueues and instance locks
LMON: Issues heartbeats and performs recovery
LMSn: Processes global cache fusion requests
LCK0: Is involved in library and row cache locking
s a
RCBG: Processes Global Result Cache invalidations
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eare other
n s feRAC backgrounds, but this lesson concentrates only on Global
e o
Note:tThere
tra
r n-Control.
v
Concurrency
no Control File to Memory Service (ACMS): In a RAC environment, the ACMS per-
E Atomic
instance process is an agent that contributes to ensuring that a distributed SGA memory
update is either globally committed on success or globally aborted if a failure occurs.
Global Enqueue Service Monitor (LMON): The LMON process monitors global
enqueues and resources across the cluster and performs global enqueue recovery
operations.
Global Enqueue Service Daemon (LMD): The LMD process manages incoming remote
resource requests within each instance.
Global Cache Service Process (LMS): The LMS process maintains records of the data
file statuses and each cached block by recording information in the GRD. The LMS
process also controls the flow of messages to remote instances and manages global
data block access and transmits block images between the buffer caches of different
instances. This processing is part of the cache fusion feature.

Instance Enqueue Process (LCK0): The LCK0 process manages noncache fusion
resource requests such as library and row cache requests.
Global Cache/Enqueue Service Heartbeat Monitor (LMHB): LMHB monitors LMON,
LMD, and LMSn processes to ensure that they are running normally without blocking or
spinning.
Result Cache Background Process (RCBG): This process is used for handling
invalidation and other messages generated by server processes attached to other
instances in Oracle RAC.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Global Resource Access Coordination
There are two types of global resource coordination.

Global enqueue management, which is used for:

Global enqueues
Global instance locks
Global buffer cache, which:
Is also known as cache fusion or global cache
Is a logical buffer cache spanning all instances s a
) ha
Coordinates access to block images in the global cache m
o
c e
Supports Parallel Query across the global caches y s d
u n i u i

br ent G
@
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E are s e
fused
Globalto
r enqueues
t r a n to control access to resources, where the owner(s), waiter(s) if
ve or converter(s)
Eany, - if any, or both, may be sessions in the same or different instances. Some
non
global enqueues serve the same purpose they would serve in a single instance. For example,
table manipulation (TM) enqueues, transaction enqueues (TX), control file enqueues (CF), high
watermark enqueues (HW), sequence cache replenishment (SQ) and redo thread enqueues
(RT) all serve the same purpose as they would in a single instance. However, there are
master and shadow metadata structures as described earlier in this lesson in the GRD, and the
mastering instance will keep track of the waiters and converters.
Instance locks are enqueues that represent resources in the row cache or library cache
protected within each instance by pins, mutexes, or latches. For cross-instance concurrency
control, an enqueue is used, the owner(s) of which is or are the instance(s) that is or are
currently the source of truth with regard to the current state of that resource. The LCK0
process acts as the owner, waiter, or converter of the enqueue as a proxy process
representing the instance. These enqueues are known as instance locks.

Global Enqueues
Processing starts in the requesting instance as follows:

1. A global enqueue request is made by a session.

2. The request is passed to LMD0 in the requesting instance.
3. The foreground waits for the request on event.
4. LMD0 determines the mastering instance.
5. LMD0 forwards the request to the mastering instance if required.
6. The mastering instance adds a new master resource if required.
s a
Process is made an owner, waiter, or converter as appropriate.
) ha
o m
c in
Once the resource can be granted to the requestor, s LMD0
is uide
y
the mastering instance notifies LMD0 in the requesting
u n
instance. b r nt G
7. When the resource is available, the n t o@ tudiseposted by
foreground
e S
cim this
LMD0 in the requesting instance.
as use
n
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
If requesting n fe instances are the same, then LMD0 need not forward the request
Eand mastering
s
e o tra LMD0 in the mastering instance notifies LMD0 in the requesting instance
t interconnect.
rthe
over
v
Ewhethernthe n -
o resource is available to the requestor immediately.
If a dequeue request is passed to the mastering instance, then LMD0 notifies the LMD0
processes for any waiters or converters that need resuming and they are posted by the LMD0
in their own instances.

Instance Locks
Instance locks are used to represent which instance(s) has

(have) control over an instance-wide structure:

Row cache entries
Library cache entries
Result cache entries
The owner, waiter, or converter on an instance lock is the
LCK0 process. a
a s
As long as the local LCK0 process in an instance owns the
m )h
lock on a specific resource, any session in that instance co can
use the cached metadata, because it is considered i s s
y current.
i d e
r u n Gu
If the local instance does not own the b then
lock, e n ta request
@ tud waits on DFS
must be made for the lock and the
e ntoforeground S
Lock Handle wait event.cim
s e t his
n na us
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Instance n E s fe structures but the scope is different. LCK0 acts as the owner or
locks usenenqueue
e o
rtfor such a
trsituations.
v
waiter n - The owner of an instance lock represents the instance having
no to access the related entity in the row cache or library cache. Assuming that an
Epermission
instance owns the instance lock, then the usual latches or pins or mutexes provide
concurrency control within the instance as usual.

Global Cache Management: Overview
Global cache management provides:

A concurrency mechanism for multiple buffer caches

An optimization of block access for reads
An optimization of writes for dirty buffers
A mechanism to optimize parallel queries
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Econtroln s fe situations where the same block has multiple images in the
e o
Concurrency handles
trabuffer caches.
rtor different
v
same n -
no in physical I/O is achieved by having a global view of buffer cache resources and
EReduction
potentially satisfying I/O requests from any instances cache, rather than reading from disk or
writing the same buffer multiple times to disk from different buffer caches
Parallel queries may result in caching parts of a table in each buffer cache and using cache
fusion to avoid repeatedly doing direct reads from disk. This is known as In Memory Parallel
Query. Parts of the table are cached in separate buffer caches, rather than having the blocks
cached multiple times in different caches due to cache fusion block transfer. The parallel
execution servers in the different instances serve results over the interconnect for the part of
the table in their respective instance caches.

Global Cache Management Components
The LMSn processes

Buffers
Buffer headers
Global Cache Master Resources
Global Cache Shadow Resources
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Multiple n Eprocesses
n s femay be used by Oracle RAC depending on the workload size. There
e o
rt headers
LMS
tra for each buffer in each buffer cache. There may be multiple block images
v
are buffer n -
o database block in the same or different buffer caches.
Efor the nsame
The mastering instance for a specific database block will have master metadata in the GRD,
maintained by LMSn, describing all block images, specifying the instance and state of each
image. The shadow metadata in the GRD, is maintained by LMSn in the same instance,
describing the state of each image

Global Cache Buffer States
Buffer states are visible in V$BH.STATUS.

Important buffer states are:

Shared Current: SCUR
Exclusive Current: XCUR
Consistent Read: CR
Built in the Instance
Sent by cache fusion
s a
Converted from SCUR or PI )ha
Past Image: PI
m
co
s
Converted from XCUR
u n isy uide
Not normally written b r nt G

Converted to CR after later XCUR n t o@ is twritten

image u de

i m e is S

a c
Multiple PI images maysexist for same
e th block in different buffer
caches.
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n Ev states s f eforr cache fusion in V$BH.STATUS are:
e r to -tran
Important buffer
v
EShared on may
nCurrent:
more instances
The buffer contains a block image that matches the one on disk. One or
have images for the same block in SCUR state. After an instance has one
in this state, cache fusion is used if another instance reads the same block for read purposes.
Exclusive Current: The buffer contains a block image that is about to be updated, or has
been updated. It may or may not have been written by the database writer. Only one instance
may have an XCUR image for a block
Consistent Read: The buffer contains a block image that is consistent with an earlier point in
time. This image may have been created in the same way as in single-instance databases,
but copying a block into an available buffer and using undo to roll back the changes in order
to create the older image. It may also get created by converting a block image from SCUR or
PI.
Past Image: The buffer contains a block image that was XCUR but then shipped to another
instance using cache fusion. A later image of this block now exists in another buffer cache.
Once DBWn writes the later image to disk from the other instance, the PI image becomes a CR
image.

Global Cache Management Scenarios
for Single Block Reads
There are several scenarios for single block reads:
Read from Disk

Read Read
Read Write
Write Write
Write Read
s a
Write to Disk
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe when an I/O request occurs for a block that has no image in any
EDisk: nOccurs
e o
Readtfrom
r cachen-tra
v
buffer
o Occurs when a block image exists in at least one buffer cache in shared
ERead nRead:
current state (SCUR), and another instance wishes to access the block for read
Read Write: Occurs when a block image exists in at least one buffer cache in shared
current state (SCUR), and another instance wishes to access the block for update (XCUR)
Write Write: Occurs when a block image exists in one buffer cache in exclusive current
state (XCUR), and another instance wishes to access the same block for write in exclusive
current state (XCUR)
Write Read: Occurs when a block image exists in one buffer cache in exclusive current
state (XCUR), and another instance wishes to access the block for read. The instance doing
the read may get it in CR or in SCUR as will be described later.
Write to Disk: Occurs when DBWn writes a dirty buffer to disk. If the block was modified in
multiple instances, then only the latest image will be written. This image will be (XCUR). All the
older dirty images for the same block will be past images (PI).

Global Cache Scenarios: Overview
Instance A Instance B
Instance C
s a
) ha
o m
y s c e
u n is uid
b r n t G
Resource
t o@ tud1008 e
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E slides,
n s fea four-instance database is used to demonstrate various scenarios
e o
In thetfollowing
r cache trafusion processing for a single block.
v
involving n -
no instance for this block is D, and the block in the database is at SCN 1008 on
EThe mastering
disk but with no images in any buffer cache.
Note that the status listed in the boxes representing the instances in the status column from
V$BH and not the GRD metadata, which has its own statuses.

Scenario 1: Read From Disk
Request to Instance C
obtain a shared
resource on C s a
) ha
1 o m
y s c e
u n is uid
b r n t G
Resource
t o@ tud1008 e
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Scenario
n s fe
e o
Read from Disk
rt n-tra
v
E noC wishes to read a block in shared mode. The foreground process passes the
Instance
request to a local LMS process, which passes it over the interconnect to an LMS process in the
mastering instance for this block.
At the moment, there is no buffer in the buffer cache on instance C, nor in any other instance,
containing a block image for the block.

Instance C
SCUR
s a
) ha
1 o m
y s c e
is id
The request is granted,run t Gu
Resource and the requesting @ b en
master
2
instance is e nt o
informed. S tud1008
s c im this
Instance D n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
The LMS n E insthe
process n
femastering instance updates master metadata in instance D and
e o trato LMS on the requesting instance C, which creates a shadow metadata
rt a grant
issues
v n -
Eresourcenoand notifies the foreground.
The buffer header status column shows SCUR for shared current.

Instance C
SCUR s a
) ha
1 o m
y s c e
u n is uid
b r n t G
Resource 2 Read nto
@ tude
master
m
request
i e is S 1008
a s c e th 3
Instance D
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n Ev process s f er
r t o
The foreground
t r a n on instance C then obtains an available buffer, and issues the I/O
ve on-
Erequest.n

Instance C
SCUR Block image a
4 delivered a s
1008 m )h
1
s co
u n isy uide
b r nt G
Resource 2 t o@ tud1008 e
n
me thi3s S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
The statusn E block
of this n s feimage after I/O completes will be SCUR even though, currently,
e o traonly instance with an image of this block.
rt Cnis-the
v
instance
o
EIf otherninstances wish to read the same block for read access, then the block image may be
sent to them via cache fusion, and two or more instances may then have images of the same
block in their buffer caches. If this occurs, then they will all be SCUR.
Note that the SCN for the block image in instance C matches the SCN of the block on disk.

Scenario 2: Read-Write Cache Fusion
Instance A Request to obtain Instance B

exclusive resource on
instance B
Instance C
SCUR s a
)ha
1008 m
s co
u n isy uide
b r nt G
Resource
t o@ tud1008 e
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe Scenario
ECachensFusion
e o
Read-Write
rt n-tra
v
Now
no B wishes to read the same block. The server process passes the request to a
E instance
local LMS process, which passes it over the interconnect to an LMS process in the mastering
instance for this block.
At the moment, there is no buffer in the buffer cache on instance B containing a block image
for the block, and there is no shadow metadata for this block in the GRD in instance B.

Instance C
SCUR s a
)ha
1008 m
s co
u n isy uide
b r nt G
Resource Instruction to transfer @ e
master
2
the block to Befor nto Stud1008
exclusive c m this
iaccess
n s
a use
Instance D
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
The mastering
E instance
n s fehas master metadata, which indicates that instance C has an SCUR
e o tra that may be sent to the requesting instance B.
rt of then-block
v
image
ESo LMSnono instance D sends a request to LMS on instance C, to send an image of the block to
instance B over the interconnect.
Note that the instance chosen to serve the block image is determined internally if two or more
instances have SCUR images for the same block.

1008
1
Instance C sends block
Instance C image to instance B
SCUR>CR
s a
3
)ha
1008 m
s co
u n isy uide
b r nt G
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fean image of the block to instance B over the interconnect. The shadow
e o
Instance C now
rt innthe sends
traGRD in instance C is changed to CR, to reflect that the block image is no
metadata
v o -
Elonger nshared.
Note that if two or more instances have SCUR images for the same block, then they will all be
messaged to downgrade them to CR by LMS on the mastering instance D.
Both images have show an SCN of 1008, but this will change when the block image is updated
in instance B.

Resource assumption
Instance A 4 Instance B
and status message

XCUR
1009
1
Instance C
CR
s a
3
)ha
1008 m
s co
u n isy uide
b r nt G
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Bnnow
LMS onoinstance s fesends a status message to mastering instance D, over the
e rt n-The
interconnect.
v tra shadow metadata in the GRD in instance B reflects that the block image is
o
Eowned nexclusive.
The mastering instance can update the master metadata because there will be two images of
the block in different instances in different states.
The image in instance B will be exclusive from the point of view of the GRD and that in
instance C will be null, because CR images do not represent blocks that reside on disk
currently or that will get written to disk. They are only used for read-consistency purposes.
LMS on instance B now posts the foreground process, and the block is updated. The SCN is
now 1009 in instance B but is still 1008 in instance C.

Scenario 3: Write-Write Cache Fusion
XCUR
1009
1
Instance C
Request to obtain CR a
resource in a s
exclusive mode 1008 m )h
s co
u n isy uide
b r nt G
Resource
t o@ tud1008 e
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe Scenario
ECachensFusion
e o
Write-Write
rt n-tra
v
E a nforeground
Now o process on instance A requests the local LMS on instance A to access the
block for write.
LMS on instance A sends a request to LMS on mastering instance D for access to the block in
exclusive mode.

XCUR
1009
1
Instance C
CR
s a
)ha
1008 m
s co
u n isy uide
b r nt G
Resource 2 Instruction to transfer t o@ tud1008 e
master exclusive resource n
e to is S
instancesA c i m th
Instance D a
n o us e
r t o n t
v e s e
n (e 2012,
r t o n le
e a b
n
LMS onomastering s fer D examines the mastering metadata in the GRD, which indicates
Ev ninstance
e rt n-Btrhasa exclusive ownership of the resource for this block.
v
that instance
o
ELMS onnmastering instance D then sends a request to LMS on instance B, requesting that the
XCUR image of the block be sent to LMS on instance A, but to keep the past image (PI) of the
block.
Instance B must flush redo buffers, if not already written, that contain the change vectors
describing the changes made to this block, before sending the block image over the
interconnect. This is not reflected in the slide, which only shows the cache fusion. Sending the
image to another instance is treated, for recovery purposes, in the same way as writing the
block to disk, and the log buffer must be written to permit recovery to work.

Exclusive-keep XCUR>PI
3 copy of buffer
1009 1009
1
Instance C
CR
s a
)ha
1008 m
s co
u n isy uide
b r nt G
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Bnthen
LMS onoinstance s fesends the block image to LMS on instance A, but retains the PI image
e
invits
t buffer
rown n - tracache. The past image of the block is held until one of the following occurs:
o version of the of the block is written by DBWn from an instance, at which time the
EA later nXCUR
PI image becomes a CR image.
LMS instructs DBWn on instance B to write the PI image, due to instance recovery. After
recovery is finished, the PI image becomes a CR image.
Note that at the moment, the SCN for the block images in instance A and B are both 1009, but
this will change when the transaction in instance A updates the block image

XCUR PI
3
1013 1009
1 4
Instance C
Resource
assumption
CR
s a
and status )ha
1008 m
message
s co
u n isy uide
b r nt G
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Annow
LMS onoinstance s fesends a status message to the mastering instance D over the
v e rt n-tra
interconnect.
EThe LMSnoon instance A updates the shadow metadata and posts the foreground process. The
status is set to XCUR in the buffer header in instance A.
The foreground updates the block and the SCN changes to 1013.
Note that this process may be repeated in a daisy-chain fashion, resulting in an XCUR image
in one instance and two or more PI images for the same block in other instances at different
points in time.

Scenario 4: Write-Read Cache Fusion
XCUR PI
1013 1009
Instance C
Request to obtain
resource in s a
shared mode 1
) ha
o m
y s c e
u n is uid
b r n t G
Resource
t o@ tud1008 e
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe Scenario
ECachensFusion
e o
Write-Read
rt n-tra
v
E a nforeground
Now o process on instance C requests local LMS on instance C to access the
block for read. LMS on instance C sends a request to LMS on mastering instance D for access
to the block in shared mode.
Note that the PI image in instance B may or may not have aged out at this point. This
depends on whether DBWn in instance A has written the XCUR image to disk. If the XCUR
image for the block in instance A, or a later XCUR image in any instance for the same block is
written by DBWn, then the PI block image becomes a CR block image, and it might age out of
the cache due to pressure on the replacement list for available buffers. This transition is
controlled by the LMSn processes in the affected instances communicating with the LMSn
process on the resource mastering instance.
Note: To simplify the slide, the CR image in instance C has aged out of the buffer cache and,
therefore, is not present.

XCUR PI
1013 1009
2
Instance C
Instruction
to transfer s a
shared 1
)ha
resource m
co
s
to C
u n isy uide
b r nt G
Resource
t o@ tud1008 e
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E ninstance
LMS onomastering s fe D examines the mastering metadata in the GRD, which indicates
e
that rt n-Atrhas
instance
a exclusive ownership of the resource for this block.
v o
ELMS onnmastering instance D then sends a request to LMS on instance A, requesting that an
image of the block satisfying the SCN of the query on instance C be sent to instance C.

XCUR Shared-keep PI
3
copy of buffer
1013 1009
2
Instance C
s a
1
)ha
1013 m
s co
u n isy uide
b r nt G
Resource
t o@ tud1008 e
n
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Normally,n E instance
LMS on n s fe A builds a CR image by using undo and sends it to instance C. The
e o tra XCUR in the buffer cache on instance A, even if the XCUR image satisfies
rtthennremains
buffer
v -
no
Ethe read-consistency requirements. This saves downgrading to SCUR and having to obtain the
image in XCUR again if another update occurs in instance A.
But if the block also has PI images, caused by multiple updates on the same block from
different instances, it may require multiple rollbacks of undo to create a consistent read image
that satisfies the request. If so, then another instance might be requested to build the CR
image by the mastering instance.
Please note that in this slide, the SCN for the block images in instance A and C are both 1013,
because in this example the XCUR image did not require rolling back to an earlier version of
the block.
Note also that the old CR image of this block could still be in the buffer cache of instance C, if
it had not aged out yet due to pressure on the replacement list.

XCUR or SCUR PI
3
1013 1009
2
Instance C
CR or SCUR
s a
1
) ha
1013 m
o
c e
y s
Resource u n is uid
b r n t G
Resource assumption e
4 information n t o@ tud1008
me this S
master
c i
Instance D n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Cnnow
LMS onoinstance s fesends a status message to mastering instance D over the
v e
interconnect, tra updates the mastering metadata.
rt n-which
o
ELMS onninstance C updates the shadowing metadata and posts the foreground process.
If no downgrade occurs, then the status is set to CR in the buffer header in instance C and
remains XCUR in instance A.
Note that CR images may age out of buffer caches, and if the block remains XCUR in instance
A, repeated cache fusion requests may occur, resulting in repeated construction and shipping
of CR images to other instances for the same block. When enough such requests occur, LMS
on the mastering instance D would request that instance A downgrade the block ownership to
SCUR and ship the image to the other instance(s). The block would also be SCUR in the other
instances.

Global Cache Management Scenarios for
Multi-Block Reads
When multi-block read requests occur:
The instance doing the I/O must acquire resources for each
block in the correct state
This is done by LMSn coordination from the requesting
instance to the LMSn on the mastering instance(s)
Different blocks in the same multi-block read may have
different mastering instances
s a
) ha
Dynamic remastering, described earlier, may help reduce the
performance overhead om c e
s
isy
There are several scenarios for multi-block reads:
u n u id
No resource masters exist for any block.
br ent G
@ tudSCUR.
Resource masters for some block(s)
e nto allSare
Resource masters for some s c im block(s)
e t hissome are XCUR.
n na us
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Emasters
n s ffore any block in a particular multi-block read request: In this case, a
e o
No resource
rt is madetrato the specific mastering instance for each block in the multi-block read and
v
request n -
no granted permission by LMSn, the server process does the multi-block read from
Eafter being
disk.
Resource masters exist for at least one block in a particular multi-block read request, but it or
they are Shared Current (SCUR): This means that the block has not been modified. In this
case, a request is made to the specific mastering instance for each block in the multi-block
read and, after being granted, the processing reads from disk.
Resource Masters exist for at least one block in a particular multi-block read request, but at
least one is Exclusive Current (XCUR) and, therefore, a newer version may exist in a buffer
cache than on disk. In this case, a request is made to the specific mastering instance for each
block in the multi-block read and, after being granted, the XCUR images are transferred by
cache fusion, as described earlier, and the remaining images are read from disk in smaller
multi-block reads.

Useful Global Resource Management Views
GV$SESSION_WAIT
GV$SYSSTAT
GV$GES_STATISTICS
V$RESOURCE_LIMIT
V$BH
V$CR_BLOCK_SERVER
V$CURRENT_BLOCK_SERVER s a
) ha
V$INSTANCE_CACHE_TRANSFER o m
y s c e
V$DYNAMIC_REMASTER_STATS
u n is uid
V$GCSPFMASTER_INFO b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Quiz
Which statement about the Global Resource Directory is not

true?
a. Resource metadata is held in the Global Resource
Directory (GRD).
b. An object under global concurrency control is called an
asset.
c. Global enqueue resources are used for enqueues and s a
locks. ) ha
o m
d. Global cache resources are used for buffer cache y s ccontrol.
e
i s
n Gofu each i d
r
b ent u
e. The GRD is distributed among all active instances
database or ASM environment. to @ tud
i m en is S
a s c e th
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n E v
s f er
Answer:
e r tob -tran
v
nonb is incorrect.
EStatement

Summary

The need for global concurrency control

Global Resource Directory
How global resources are managed
RAC global resource access coordination
Global enqueue and instance lock management
Global buffer cache management s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non
RAC Database Monitoring and Tuning
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives

Determine RAC-specific tuning components

Determine RAC-specific wait events, global enqueues, and
system statistics
Implement the most common RAC tuning tips
Use the Cluster Database Performance pages
Use the Automatic Workload Repository (AWR) in RAC has
a
m )
Use Automatic Database Diagnostic Monitor (ADDM) o in
y s c e
RAC is id
r u n Gu
@ b ent
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

CPU and Wait Time Tuning Dimensions
CPU
time
Possibly Scalable
needs SQL application
tuning
s a
)ha
Scalable Needs No gain achieved
m
application instance/RAC s co
by adding
tuning
u n
y
isCPUs/nodes
u ide
b r nt G
n t o@ Wait t u de
c i me this Stime
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe it is important that you compare the CPU time with the wait time of
Eyournsystem,
e o
Whenttuning
rsystem. tra
v
your n -Comparing CPU time with wait time helps to determine how much of the
Eresponsenotime is spent on useful work and how much on waiting for resources potentially held
by other processes.
As a general rule, the systems where CPU time is dominant usually need less tuning than the
ones where wait time is dominant. Alternatively, heavy CPU usage can be caused by badly
written SQL statements.
Although the proportion of CPU time to wait time always tends to decrease as load on the
system increases, steep increases in wait time are a sign of contention and must be
addressed for good scalability.
Adding more CPUs to a node, or nodes to a cluster, would provide very limited benefit under
contention. Conversely, a system where the proportion of CPU time to wait time does not
decrease significantly as load increases can scale better, and would most likely benefit from
adding CPUs or Real Application Clusters (RAC) instances if needed.
Note: Automatic Workload Repository (AWR) reports display CPU time together with wait
time in the Top 5 Timed Events section, if the CPU time portion is among the top five events.

RAC-Specific Tuning
Tune for a single instance first.

Tune for RAC:

Instance recovery
Interconnect traffic
Point of serialization can be exacerbated.
RAC-reactive tuning tools:
Specific wait events Certain combinations
s a
are characteristic of
) ha
System and enqueue statistics well-known tuning cases. m
Enterprise Manager performance pages
o
c e
y s
Statspack and AWR reports u n is uid
r G b ent
@
RAC-proactive tuning tools:
e nto Stud
AWR snapshots
s c im this
ADDM reports n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe tuning areas for RAC, such as instance recovery and interconnect
E arenspecific
Although
e o there a benefits by tuning your system like a single-instance system. At least, this
rt you nget-trmost
v
traffic,
E no
must be your starting point.
Obviously, if you have serialization issues in a single-instance environment, these may be
exacerbated with RAC.
As shown in the slide, you have basically the same tuning tools with RAC as with a single-
instance system. However, certain combinations of specific wait events and statistics are well-
known RAC tuning cases.
In this lesson, you see some of those specific combinations, as well as the RAC-specific
information that you can get from the Enterprise Manager performance pages, and Statspack
and AWR reports. Finally, you see the RAC-specific information that you can get from the
Automatic Database Diagnostic Monitor (ADDM).

Analyzing Cache Fusion Impact in RAC
The cost of block access and cache coherency is

represented by:
Global Cache Services statistics
Global Cache Services wait events
The response time for cache fusion transfers is determined
by:
Overhead of the physical interconnect components
s a
IPC protocol ) ha
o m
GCS protocol sc
The response time is not generally affected u nby diskuI/Oisy ide
r
b en t G
factors. t o @ ud t
e n S
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eaccessing
n s feblocks in the global cache and maintaining cache coherency is
e o
The effect of
rt n-by: tra
v
represented
E The noGlobal Cache Services statistics for current and cr blocksfor example, gc current
blocks received, gc cr blocks received, and so on
The Global Cache Services wait events for gc current block 3-way, gc cr grant 2-way,
and so on
The response time for cache fusion transfers is determined by the messaging time and
processing time imposed by the physical interconnect components, the IPC protocol, and the
GCS protocol. It is not affected by disk input/output (I/O) factors other than occasional log
writes. The cache fusion protocol does not require I/O to data files in order to guarantee
cache coherency, and RAC inherently does not cause any more I/O to disk than a
nonclustered instance.

Typical Latencies for RAC Operations
AWR Report Latency Name Lower Typical Upper
Bound Bound
Average time to process cr block request 0.1 1 10

Avg global cache cr block receive time (ms) 0.3 4 12
Average time to process current block request 0.1 3 23
Avg global cache current block receive time(ms) 0.3 8 30
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E report,
n s fethere is a table in the RAC Statistics section containing average times
In a RAC
e o AWR
ra Global Cache Services and Global Enqueue Services operations. This
rt nfor-tsome
v
(latencies)
o in the slide and is called Global Cache and Enqueue Services: Workload
Etable isnshown
Characteristics. Those latencies should be monitored over time, and significant increases in
their values should be investigated. The table presents some typical values, based on
empirical observations. Factors that may cause variations to those latencies include:
Utilization of the IPC protocol. User-mode IPC protocols are faster, but only Tru64s
RDG is recommended for use.
Scheduling delays, when the system is under high CPU utilization
Log flushes for current blocks served
Other RAC latencies in AWR reports are mostly derived from V$GES_STATISTICS and may
be useful for debugging purposes, but do not require frequent monitoring.
Note: The time to process consistent read (CR) block request in the cache corresponds to
(build time + flush time + send time), and the time to process current block
request in the cache corresponds to (pin time + flush time + send time).

Wait Events for RAC
Wait events help to analyze what sessions are waiting for.

Wait times are attributed to events that reflect the outcome

of a request:
Placeholders while waiting
Precise events after waiting
Global cache waits are summarized in a broader category
called Cluster Wait Class. s a
) ha
These wait events are used in ADDM to enable cache m
c o
fusion diagnostics. ys e is uid
r u n G
b n t
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E sessions
n s fe are waiting for is an important method to determine where time is
e o
Analyzing what
rt In RAC, a wait time is attributed to an event that reflects the exact outcome of a
trthe
v
spent. n -
E no
request. For example, when a session on an instance is looking for a block in the global
cache, it does not know whether it will receive the data cached by another instance or
whether it will receive a message to read from disk. The wait events for the global cache
convey precise information and wait for global cache blocks or messages. They are mainly
categorized by the following:
Summarized in a broader category called Cluster Wait Class
Temporarily represented by a placeholder event that is active while waiting for a block
Attributed to precise events when the outcome of the request is known
The wait events for RAC convey information valuable for performance analysis. They are
used in ADDM to enable precise diagnostics of the impact of cache fusion.

Wait Event Views
Total waits for an event V$SYSTEM_EVENT

Waits for a wait event class V$SESSION_WAIT_CLASS

by a session
Waits for an event by a session V$SESSION_EVENT
Activity of recent active sessions V$ACTIVE_SESSION_HISTORY a

a s
Last 10 wait events m )h
for each active session
V$SESSION_WAIT_HISTORY
s co
u n isy uide
Events for which b r nt G
active sessions are waiting
n t o@ tude
V$SESSION_WAIT
i m e is S
Identify SQL statements impacted
a s c e th V$SQLSTATS
by interconnect latencies
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n Evsome s f erto acquire resources because of the total path length and latency for
When itotakes
e r t processes
requests, - t r an sleep to avoid spinning for indeterminate periods of time. When the
time
v
Eprocess on to wait, it wakes up either after a specified timer value expires (timeout) or
n decides
when the event it is waiting for occurs and the process is posted. The wait events are
recorded and aggregated in the views shown in the slide. The first three are aggregations of
wait times, timeouts, and the number of times waited for a particular event, whereas the rest
enable the monitoring of waiting sessions in real time, including a history of recent events
waited for.
The individual events distinguish themselves by their names and the parameters that they
assume. For most of the global cache wait events, the parameters include file number, block
number, the block class, and access mode dispositions, such as mode held and requested.
The wait times for events presented and aggregated in these views are very useful when
debugging response time performance issues. Note that the time waited is cumulative, and
that the event with the highest score is not necessarily a problem. However, if the available
CPU power cannot be maximized, or response times for an application are too high, the top
wait events provide valuable performance diagnostics.
Note: Use the CLUSTER_WAIT_TIME column in V$SQLSTATS to identify SQL statements
impacted by interconnect latencies, or run an ADDM report on the corresponding AWR
snapshot.

Global Cache Wait Events: Overview
Just requested gc [current/cr] [multiblock] request

(placeholder)
gc [current/cr] block [2/3]-way gc [current/cr] block busy

Received after two or three network hops, Received but not sent immediately
immediately after request
gc [current/cr] grant 2-way gc current grant busy

Not received and not mastered locally. Not received and not mastered locally.
s a
Grant received immediately. Grant received with delay.
)ha
co m
s
gc [current/cr] [block/grant] congested
isy uide
gc [current/cr] [failure/retry]
u n
Block or grant received with delay
b r because
Not received
n t G
of failure
because of CPU or memory lack @
to Stud e
e n
s c im this
gc buffer busy
n athan buffer
Block arrival
u s etime
r t o nless
t o pin time
( e ve ense
o n n e2012,
Copyright
v e rt rab
The mainn E nsfewait events are described briefly in the slide:
global cache
o
rt n-tra
ve gc current/cr These wait events are relevant only while a gc request for
E ancroblock or currentrequest:buffer is in progress. They act as placeholders until the request
completes.
gc [current/cr] block [2/3]-way: A current or cr block is requested and
received after two or three network hops. The request is processed immediately; the
block is not busy or congested.
gc [current/cr] block busy: A current or cr block is requested and received, but
is not sent immediately by LMS because some special condition that delayed the
sending was found.
gc [current/cr] grant 2-way: A current or cr block is requested and a grant
message received. The grant is given without any significant delays. If the block is not in
its local cache, a current or cr grant is followed by a disk read on the requesting
instance.
gc current grant busy: A current block is requested and a grant message
received. The busy hint implies that the request is blocked because others are ahead of
it or it cannot be handled immediately.
Note: For dynamic remastering, two events are of most importance: gc remaster and gc
quiesce. They can be symptoms of the impact of remastering on the running processes.

gc [current/cr] [block/grant] congested: A current or cr block is requested
and a block or grant message received. The congested hint implies that the request
spent more than 1 ms in internal queues.
gc [current/cr] [failure/retry]: A block is requested and a failure status
received or some other exceptional event has occurred.
gc buffer busy: If the time between buffer accesses becomes less than the time the
buffer is pinned in memory, the buffer containing a block is said to become busy and as
a result interested users may have to wait for it to be unpinned.
Note: For more information, refer to Oracle Database Reference.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Global Enqueue Waits
Enqueues are synchronous.

Enqueues are global resources in RAC.

The most frequent waits are for:
TX TM
s a
) ha
US HW o m
y s c e
u n is uid
r t G
TA SQ @b n
n t o t u de
i m e is S
The waits may constituteaserious s c e th
serialization points.
n n u s
r t o t o
( e ve ense
o n n e2012,
Copyright
v e rt rab
n
An enqueue
Ewait isnsnotfeRAC specific, but involves a global lock operation when RAC is
e o
rt Most trofathe global requests for enqueues are synchronous, and foreground
enabled.
v n -
no wait for them. Therefore, contention on enqueues in RAC is more visible than in
Eprocesses
single-instance environments. Most waits for enqueues occur for enqueues of the following
types:
TX: Transaction enqueue; used for transaction demarcation and tracking
TM: Table or partition enqueue; used to protect table definitions during DML operations
HW: High-water mark enqueue; acquired to synchronize a new block operation
SQ: Sequence enqueue; used to serialize incrementing of an Oracle sequence number
US: Undo segment enqueue; mainly used by the Automatic Undo Management (AUM)
feature
TA: Enqueue used mainly for transaction recovery as part of instance recovery
In all of the preceding cases, the waits are synchronous and may constitute serious
serialization points that can be exacerbated in a RAC environment.
Note: The enqueue wait events specify the resource name and a reason for the waitfor
example, TX Enqueue index block split. This makes diagnostics of enqueue waits easier.

Session and System Statistics
Use V$SYSSTAT to characterize the workload.

Use V$SESSTAT to monitor important sessions.

V$SEGMENT_STATISTICS includes RAC statistics.
RAC-relevant statistic groups are:
Global Cache Service statistics
Global Enqueue Service statistics
Statistics for messages sent s a
) ha
V$ENQUEUE_STATISTICS determines the enqueue o m
with
the highest impact. y s c e
u n is uid
V$INSTANCE_CACHE_TRANSFER breaks b
r down n t G
GCS
statistics into block classes. n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
Using system n febased on V$SYSSTAT enables characterization of the database
E statistics
s
e o
rt based raaverages. It is the basis for many metrics and ratios used in various tools
ton
v
activity n -
no such as AWR, Statspack, and Database Control.
Eand methods,
In order to drill down to individual sessions or groups of sessions, V$SESSTAT is useful when
the important session identifiers to monitor are known. Its usefulness is enhanced if an
application fills in the MODULE and ACTION columns in V$SESSION.
V$SEGMENT_STATISTICS is useful for RAC because it also tracks the number of CR and
current blocks received by the object.
The RAC-relevant statistics can be grouped into:
Global Cache Service statistics: gc cr blocks received, gc cr block receive time, and
so on
Global Enqueue Service statistics: global enqueue gets, and so on
Statistics for messages sent: gcs messages sent and ges messages sent
V$ENQUEUE_STATISTICS can be queried to determine which enqueue has the highest
impact on database service times and, eventually, response times.
V$INSTANCE_CACHE_TRANSFER indicates how many current and CR blocks per block class
are received from each instance, including how many transfers incurred a delay.
Note: For more information about statistics, refer to Oracle Database Reference.

Most Common RAC Tuning Tips
Application tuning is often the most beneficial!

Reduce long full-table scans in OLTP systems.

Use Automatic Segment Space Management (ASSM).
Increase sequence caches.
Use partitioning to reduce interinstance traffic.
Avoid unnecessary parsing.
s a
Minimize locking usage.
) ha
Remove unselective indexes. o m
y s c e
Configure interconnect properly. is id
un Gu br ent
@
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
In anyto n E system,
database n s fe RAC or single-instance, the most significant performance gains are
v e r obtained
usually n - trafrom traditional application-tuning techniques. The benefits of those
E no
techniques are even more remarkable in a RAC database. In addition to traditional application
tuning, some of the techniques that are particularly important for RAC include the following:
Try to avoid long full-table scans to minimize GCS requests. The overhead caused by
the global CR requests in this scenario is because when queries result in local cache
misses, an attempt is first made to find the data in another cache, based on the
assumption that the chance is high that another instance has cached the block.
Automatic Segment Space Management can provide instance affinity to table blocks.
Increasing sequence caches improves instance affinity to index keys deriving their
values from sequences. That technique may result in significant performance gains for
multi-instance insert-intensive applications.
Range or list partitioning may be very effective in conjunction with data-dependent
routing, if the workload can be directed to modify a particular range of values from a
particular instance.
Hash partitioning may help to reduce buffer busy contention by making buffer access
distribution patterns sparser, enabling more buffers to be available for concurrent
access.

In RAC, library cache and row cache operations are globally coordinated. So, excessive
parsing means additional interconnect traffic. Library cache locks are heavily used, in
particular by applications that use PL/SQL or Advanced Queuing. Library cache locks
are acquired in exclusive mode whenever a package or procedure has to be recompiled.
Because transaction locks are globally coordinated, they also deserve special attention
in RAC. For example, using tables instead of Oracle sequences to generate unique
numbers is not recommended because it may cause severe contention even for a
single-instance system.
Indexes that are not selective do not improve query performance, but can degrade DML
performance. In RAC, unselective index blocks may be subject to inter-instance
contention, increasing the frequency of cache transfers for indexes belonging to insert-
intensive tables.
Always verify that you use a private network for your interconnect, and that your private
network is configured properly. Ensure that a network link is operating in full duplex
s
mode. Ensure that your network interface and Ethernet switches support MTU size of 9
a
) ha
KB. Note that a single-gigabit Ethernet interface can scale up to ten thousand 8 KB
o m
blocks per second before saturation.
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Index Block Contention: Considerations
Wait events
Index
enq: TX - index
contention
block
Split in
gc buffer busy
progress
gc current block busy
gc current split
s a
System statistics )ha
m
co
Leaf node splits
s
Branch node splits
u n isy uide
Exchange deadlocks b r nt G
n t o@ tude
me this S
gcs refuse xid
gcs ast xid c i
n as use RAC01 RAC02
ton e to
Service ITL waits
e r
v ens
( e
o n n e2012,
Copyright
v e rt rab
n E systems
In application n s fewhere the loading or batch processing of data is a dominant business
e o
rt there a be performance issues affecting response times because of the high
trmay
function,
v n -
Evolumenofodata inserted into indexes. Depending on the access frequency and the number of
processes concurrently inserting data, indexes can become hot spots and contention can be
exacerbated by:
Ordered, monotonically increasing key values in the index (right-growing trees)
Frequent leaf block splits
Low tree depth: All leaf block access goes through the root block.
A leaf or branch block split can become an important serialization point if the particular leaf
block or branch of the tree is concurrently accessed. The tables in the slide sum up the most
common symptoms associated with the splitting of index blocks, listing wait events and
statistics that are commonly elevated when index block splits are prevalent. As a general
recommendation, to alleviate the performance impact of globally hot index blocks and leaf
block splits, a more uniform, less skewed distribution of the concurrency in the index tree
should be the primary objective. This can be achieved by:
Global index hash partitioning
Increasing the sequence cache, if the key value is derived from a sequence
Using natural keys as opposed to surrogate keys
Using reverse key indexes

Oracle Sequences and Index Contention
Can contain 500 rows
150000 50001100000
s a
)ha
m
co
s
CACHE 50000 NOORDER
u n isy uide
b r nt G
RAC01 t o@ tudRAC02 e
n
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Ekey values
n s fe generated by sequences tend to be subject to leaf block contention
Indexes
e o with
rtthe inserttrarate is high. That is because the index leaf block holding the highest key
v
when
o n -
Evalue isnchanged for every row inserted, as the values are monotonically ascending. In RAC,
this may lead to a high rate of current and CR blocks transferred between nodes.
One of the simplest techniques that can be used to limit this overhead is to increase the
sequence cache, if you are using Oracle sequences. Because the difference between
sequence values generated by different instances increases, successive index block splits
tend to create instance affinity to index leaf blocks. For example, suppose that an index key
value is generated by a CACHE NOORDER sequence and each index leaf block can hold 500
rows. If the sequence cache is set to 50000, while instance 1 inserts values 1, 2, 3, and so on,
instance 2 concurrently inserts 50001, 50002, and so on. After some block splits, each
instance writes to a different part of the index tree.
So, what is the ideal value for a sequence cache to avoid inter-instance leaf index block
contention, yet minimizing possible gaps? One of the main variables to consider is the insert
rate: The higher it is, the higher must be the sequence cache. However, creating a simulation
to evaluate the gains for a specific configuration is recommended.
Note: By default, the cache value is 20. Typically, 20 is too small for the preceding example.

Undo Block Considerations
Index Changes
Reads
s a
SGA1 SGA2
)ha
m
co
Undo Undo s
u n isy uide
b r nt G
n t o@ tude
i me this S
Additional
c
n as usetraffic
interconnect
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Excessiven E nsfeshipment and contention for undo buffers usually happens when index
undo block
e o tra active transactions from multiple instances are read frequently.
rt containing
blocks
v o n
EWhen anSELECT-
statement needs to read a block with active transactions, it has to undo the
changes to create a CR version. If the active transactions in the block belong to more than
one instance, there is a need to combine local and remote undo information for the consistent
read. Depending on the amount of index blocks changed by multiple instances and the
duration of the transactions, undo block shipment may become a bottleneck.
Usually this happens in applications that read recently inserted data very frequently, but
commit infrequently. Techniques that can be used to reduce such situations include the
following:
Shorter transactions reduce the likelihood that any given index block in the cache
contains uncommitted data, thereby reducing the need to access undo information for
consistent read.
As explained earlier, increasing sequence cache sizes can reduce inter-instance
concurrent access to index leaf blocks. CR versions of index blocks modified by only
one instance can be fabricated without the need of remote undo information.
Note: In RAC, the problem is exacerbated by the fact that a subset of the undo information
has to be obtained from remote instances.

High-Water Mark Considerations
Wait events
enq: HW -
contention
gc current grant Heavy
inserts
HWM
Heavy
s a
inserts )ha
m
co
s
u n isy uide
b r nt G
New extent n t o@ tude
c i me this S
RAC01 n as use RAC02
n
rto se t o
n (e 2012,
r t o n le
e r a b
A certain n Ev nsfe
combination of wait events and statistics presents itself in applications where the
r t o t r a
ve oton-a segment. If databusiness
insertion
Efrequently
of data is a dominant function and new blocks have to be allocated
n is inserted at a high rate, new blocks may have to be made
available after unfruitful searches for free space. This has to happen while holding the high-
water mark (HWM) enqueue.
Therefore, the most common symptoms for this scenario include:
A high percentage of wait time for enq: HW contention
A high percentage of wait time for gc current grant events
The former is a consequence of the serialization on the HWM enqueue, and the latter is
because of the fact that current access to the new data blocks that need formatting is required
for the new block operation. In a RAC environment, the length of this space management
operation is proportional to the time it takes to acquire the HWM enqueue and the time it
takes to acquire global locks for all the new blocks that need formatting. This time is small
under normal circumstances because there is never any access conflict for the new blocks.
Therefore, this scenario may be observed in applications with business functions requiring a
lot of data loading, and the main recommendation to alleviate the symptoms is to define
uniform and large extent sizes for the locally managed and automatic space-managed
segments that are subject to high-volume inserts.

Concurrent Cross-Instance Calls: Considerations
Dirty
block
SGA1 SGA2
Table1 Table1
CKPT CKPT
Table2 Table2
s a
)ha
m
co
1
s
u n isy 2 uide
3 4 b r nt G
Truncate Table1 n t o@ Truncate t u de Table2
c
Cross-instance i me callthis S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nand
In dataowarehouse s fedata mart environments, it is not uncommon to see a lot of TRUNCATE
v e rt nThese
operations. - tra essentially happen on tables containing temporary data.
noenvironment, truncating tables concurrently from different instances does not scale
EIn a RAC
well, especially if, in conjunction, you are also using direct read operations such as parallel
queries.
As shown in the slide, a truncate operation requires a cross-instance call to flush dirty blocks
of the table that may be spread across instances. This constitutes a point of serialization. So,
while the first TRUNCATE command is processing, the second has to wait until the first one
completes.
There are different types of cross-instance calls. However, all use the same serialization
mechanism.
For example, the cache flush for a partitioned table with many partitions may add latency to a
corresponding parallel query. This is because each cross-instance call is serialized at the
cluster level, and one cross-instance call is needed for each partition at the start of the parallel
query for direct read purposes.

Monitoring RAC Database
and Cluster Performance
Database
Directly from Database Control and Grid Control:
View the status of each node in the cluster.

View the aggregated alert messages across
Instances
all the instances.
Review the issues that are affecting the entire cluster or
each instance.
Monitor the cluster cache coherency statistics. s a
) ha
Determine whether any of the services for the cluster o m
c
database are having availability problems. isys ide
r u n Gu
Review any outstanding Clusterware interconnect
b nt alerts.
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EEnterprise
n s feManager Database Control and Grid Control are cluster-aware and
e o
Both Oracle
traconsole to manage your cluster database. From the Cluster Database Home
rt a central
v
provide n -
nocan do all of the following:
Epage, you
View the overall system status, such as the number of nodes in the cluster and their
current status, so you do not have to access each individual database instance for
details.
View the alert messages aggregated across all the instances with lists for the source of
each alert message.
Review the issues that are affecting the entire cluster as well as those that are affecting
individual instances.
Monitor cluster cache coherency statistics to help you identify processing trends and
optimize performance for your Oracle RAC environment. Cache coherency statistics
measure how well the data in caches on multiple instances is synchronized.
Determine whether any of the services for the cluster database are having availability
problems. A service is deemed to be a problem service if it is not running on all
preferred instances, if its response time thresholds are not met, and so on.
Review any outstanding Clusterware interconnect alerts.

Cluster Database Performance Page
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EDatabase
n s fePerformance page provides a quick glimpse of the performance
e o
The Cluster
ra
rt forna-tdatabase.
v
statistics Enterprise Manager accumulates data from each instance over
noperiods of time, called collection-based data. Enterprise Manager also provides
Especified
current data from each instance, known as real-time data.
Statistics are rolled up across all the instances in the cluster database. Using the links next to
the charts, you can get more specific information and perform any of the following tasks:
Identify the causes of performance issues.
Decide whether resources need to be added or redistributed.
Tune your SQL plan and schema for better optimization.
Resolve performance issues.
The screenshot in the slide shows a partial view of the Cluster Database Performance page.
You access this page by clicking the Performance tab from the Cluster Database Home page.

Determining Cluster Host Load Average
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EHost Load
n s feAverage chart in the Cluster Database Performance page shows
e o
The Cluster
tra that are outside the database. The chart shows maximum, average, and
rt problems
v
potential n -
Eminimumnoload values for available nodes in the cluster for the previous hour.
If the load average is higher than the average of the total number of CPUs across all the hosts
in the cluster, then too many processes are waiting for CPU resources. SQL statements that
are not tuned often cause high CPU usage. Compare the load average values with the values
displayed for CPU Used in the Average Active Sessions chart. If the sessions value is low
and the load average value is high, this indicates that something else on the host, other than
your database, is consuming the CPU.
You can click any of the load value labels for the Cluster Host Load Average chart to view
more detailed information about that load value. For example, if you click the Average label,
the Hosts: Average Load page appears, displaying charts that depict the average host load
for up to four nodes in the cluster.
You can select whether the data is displayed in a summary chart, combining the data for each
node in one display, or using tile charts, where the data for each node is displayed in its own
chart. You can click Customize to change the number of tile charts displayed in each row or
the method of ordering the tile charts.

Determining Global Cache Block Access Latency
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
The Global
E s
CachenBlock
fe Access Latency chart shows the latency for each type of data block
e o
rt currenttraand consistent-read (CR) blocks. That is the elapsed time it takes to locate
request:
v n -
no consistent-read and current blocks between the buffer caches.
Eand transfer
You can click either metric for the Global Cache Block Access Latency chart to view more
detailed information about that type of cached block.
If the Global Cache Block Access Latency chart shows high latencies (high elapsed times),
this can be caused by any of the following:
A high number of requests caused by SQL statements that are not tuned
A large number of processes in the queue waiting for the CPU, or scheduling delays
Slow, busy, or faulty interconnects. In these cases, check your network connection for
dropped packets, retransmittals, or cyclic redundancy check (CRC) errors.
Concurrent read and write activity on shared data in a cluster is a frequently occurring activity.
Depending on the service requirements, this activity does not usually cause performance
problems. However, when global cache requests cause a performance problem, optimizing
SQL plans and the schema to improve the rate at which data blocks are located in the local
buffer cache, and minimizing I/O is a successful strategy for performance tuning. If the latency
for consistent-read and current block requests reaches 10 milliseconds, then see the Cluster
Cache Coherency page for more detailed information.

Determining Average Active Sessions
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EActive n s fe
r t o
The Average
t r a Sessions chart on the Cluster Database Performance page shows
ve oproblems
Epotential - inside the database. Categories, called wait classes, show how much of
n n is using a resource, such as CPU or disk I/O. Comparing CPU time with wait
the database
time helps to determine how much of the response time is consumed with useful work rather
than waiting for resources that are potentially held by other processes.
At the cluster database level, this chart shows the aggregate wait class statistics across all
the instances. For a more detailed analysis, you can click the Clipboard icon at the bottom of
the chart to view the ADDM analysis for the database for that time period.
If you click the wait class legends beside the Average Active Sessions chart, you can view
instance-level information stored in Active Sessions by Instance pages. You can use the
Wait Class action list on the Active Sessions by Instance page to view the different wait
classes. The Active Sessions by Instance pages show the service times for up to four
instances. Using the Customize button, you can select the instances that are displayed. You
can view the data for the instances separately by using tile charts, or you can combine the
data into a single summary chart.

Determining Database Throughput
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E on the n s fe
The last
r t o chart
t r a Performance page monitors the usage of various database resources.
ve theoThroughput
EClick n- tab at the top of this chart to view the Database Throughput chart.
Compare n the peaks on the Average Active Sessions chart with those on the Database
Throughput charts. If internal contention is high and throughput is low, consider tuning the
database.
The Database Throughput charts summarize any resource contention that appears in the
Average Active Sessions chart, and also show how much work the database is performing on
behalf of the users or applications. The Per Second view shows the number of transactions
compared to the number of logons, and (not shown here) the number of physical reads
compared to the redo size per second. The Per Transaction view shows the number of
physical reads compared to the redo size per transaction. Logons is the number of users that
are logged on to the database.
To obtain information at the instance level, access the Database Throughput by Instance
page by clicking one of the legends to the right of the charts. This page shows the breakdown
of the aggregated Database Throughput chart for up to four instances. You can select the
instances that are displayed. You can drill down further on the Database Throughput by
Instance page to see the sessions of an instance consuming the greatest resources. Click an
instance name legend under the chart to go to the Top Sessions subpage of the Top
Consumers page for that instance.

Determining Database Throughput
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E on then s fe
The last
r t o chart
t r a Performance page monitors the usage of various database resources.
veclicking
EBy - Instances tab at the top of this chart, you can view the Active Sessions by
nthe
Instancenochart.
The Active Sessions by Instance chart summarizes any resource contention that appears in
the Average Active Sessions chart. Using this chart, you can quickly determine how much of
the database work is being performed on each instance.
You can also obtain information at the instance level by clicking one of the legends to the right
of the chart to access the Top Sessions page. On the Top Sessions page, you can view real-
time data showing the sessions that consume the greatest system resources. In the graph in
the slide, the orac2 instance after 8:20 PM is consistently showing more active sessions than
the orac1 instance.

Accessing the Cluster Cache Coherency Page
Block
Class
Segment s a
name )ha
m
co
s
u n isy uide
Segment b r nt G
name
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Cluster
n s feCache Coherency page, click the Performance tab on the Cluster
e o
To access
rt Homethe
trapage, and click Cluster Cache Coherency in the Additional Monitoring Links
v
Database n -
Esectionnatothe bottom of the page. Alternatively, click either of the legends to the right of the
Global Cache Block Access Latency chart.
The Cluster Cache Coherency page contains summary charts for cache coherency metrics for
the cluster:
Global Cache Block Access Latency: Shows the total elapsed time, or latency, for a
block request. Click one of the legends to the right of the chart to view the average time
it takes to receive data blocks for each block type (current or CR) by instance. On the
Average Block Receive Time by Instance page, you can click an instance legend
under the chart to go to the Block Transfer for Local Instance page, where you can
identify which block classes, such as undo blocks, data blocks, and so on, are subject to
intense global cache activity. This page displays the block classes that are being
transferred, and which instances are transferring most of the blocks. Cache transfer
indicates how many current and CR blocks for each block class were received from
remote instances, including how many transfers incurred a delay (busy) or an
unexpected longer delay (congested).

Accessing the Cluster Cache Coherency Page
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n ECache n s fe Transfer Rate: Shows the total aggregated number of blocks
e t o
Global
rreceived Block
trbyaall instances in the cluster by way of an interconnect. Click one of the
v n -
no to the right of the chart to go to the Global Cache Blocks Received by Instance
E legends
page for that type of block. From there, you can click an instance legend under the chart
to go to the Segment Statistics by Instance page, where you can see which segments
are causing cache contention.
Global Cache Block Transfers and Physical Reads: Shows the percentage of logical
read operations that retrieved data from the buffer cache of other instances by way of
Direct Memory Access and from disk. It is essentially a profile of how much work is
performed in the local buffer cache, rather than the portion of remote references and
physical reads, which both have higher latencies. Click one of the legends to the right of
the chart to go to the Global Cache Block Transfers vs. Logical Reads by Instance and
Physical Reads vs. Logical Reads by Instance pages. From there, you can click an
instance legend under the chart to go to the Segment Statistics by Instance page,
where you can see which segments are causing cache contention.

Viewing the Cluster Interconnects Page
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
The Clustern n fe page is useful for monitoring the interconnect interfaces,
EInterconnects
s
e o tra
rt nconfiguration
determining
v o
Etraffic. This - issues, and identifying transfer raterelated issues including excess
n page helps to determine the load added by instances and databases on the
interconnect. Sometimes you can quickly identify interconnect delays that are due to
applications outside Oracle.
You can use this page to perform the following tasks:
View all interfaces that are configured across the cluster.
View statistics for the interfaces, such as absolute transfer rates and errors.
Determine the type of interfaces, such as private or public.
Determine whether the instance is using a public or private network.
Determine which database instance is currently using which interface.
Determine how much the instance is contributing to the transfer rate on the interface.
The Private Interconnect Transfer Rate value shows a global view of the private interconnect
traffic, which is the estimated traffic on all the private networks in the cluster. The traffic is
calculated as the summary of the input rate of all private interfaces known to the cluster.
From the Cluster Interconnects page, you can access the Hardware Details page, on which
you can get more information about all the network interfaces defined on each node of your
cluster.

Similarly, you can access the Transfer Rate metric page, which collects the internode
communication traffic of a cluster database instance. The critical and warning thresholds of
this metric are not set by default. You can set them according to the speed of your cluster
interconnects.
Note: You can query the GV$CLUSTER_INTERCONNECTS view to see information about the
private interconnect:
SQL> select * from GV$CLUSTER_INTERCONNECTS;
INST_ID NAME IP_ADDRESS IS_PUBLIC SOURCE

-------- ----- --------------- --------- -------------------------
1 eth1 192.0.2.110 NO Oracle Cluster Repository
2 eth1 192.0.2.111 NO Oracle Cluster Repository
3 eth1 192.0.2.112 NO s
Oracle Cluster Repository a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Viewing the Database Locks Page
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nLocks s fe page to determine whether multiple instances are holding locks for
Use the
e oDatabase
rt object.traThe page shows user locks, all database locks, or locks that are blocking
v
the same n -
no or applications. You can use this information to stop a session that is
Eother users
unnecessarily locking an object.
To access the Database Locks page, select Performance on the Cluster Database Home
page, and click Database Locks in the Additional Monitoring Links section at the bottom of the
Performance subpage.

AWR Snapshots in RAC
MMON Coordinator
In-memory
statistics
SYSAUX
SGA (Inst1)
AWR tables
6:00 a.m.
9:00 a.m. s a
7:00 a.m.
)ha
8:00 a.m. m
co
MMON s
isy uide
9:00 a.m.
In-memory
statistics u n
r nt G
b
n t o@ tude
SGA (Instn)
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe
E nsgenerates
r o
AWR tautomatically
t r a snapshots of the performance data once every hour and
ve the
Ecollects -
nstatistics in the workload repository. In RAC environments, each AWR snapshot
nodata
captures from all active instances within the cluster. The data for each snapshot set that
is captured for all active instances is from roughly the same point in time. In addition, the data
for each instance is stored separately and is identified with an instance identifier. For
example, the buffer_busy_wait statistic shows the number of buffer waits on each
instance. The AWR does not store data that is aggregated from across the entire cluster. That
is, the data is stored for each individual instance.
The statistics snapshots generated by the AWR can be evaluated by producing reports
displaying summary data such as load and cluster profiles based on regular statistics and wait
events gathered on each instance.
The AWR functions in a similar way as Statspack. The difference is that the AWR
automatically collects and maintains performance statistics for problem detection and self-
tuning purposes. Unlike in Statspack, in the AWR, there is only one snapshot_id per
snapshot across instances.

AWR Reports and RAC: Overview
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E statistics
n s fe in an AWR report are organized in different sections. A RAC
e o
The RAC-related
rt section traappears after the Top 5 Timed Events. This section contains:
v
statistics n -
E The nonumber of instances open at the time of the begin snapshot and the end snapshot
to indicate whether instances joined or left between the two snapshots
The Global Cache Load Profile, which essentially lists the number of blocks and
messages that are sent and received, as well as the number of fusion writes
The Global Cache Efficiency Percentages, which indicate the percentage of buffer gets
broken up into buffers received from the disk, local cache, and remote caches. Ideally,
the percentage of disk buffer access should be close to zero.
GCS and GES Workload Characteristics, which gives you an overview of the more
important numbers first. Because the global enqueue convert statistics have been
consolidated with the global enqueue get statistics, the report prints only the average
global enqueue get time. The round-trip times for CR and current block transfers follow,
as well as the individual sender-side statistics for CR and current blocks. The average
log flush times are computed by dividing the total log flush time by the number of actual
log flushes. Also, the report prints the percentage of blocks served that actually incurred
a log flush.

GCS and GES Messaging Statistics. The most important statistic here is the average
message sent queue time on ksxp, which indicates how well the IPC works. Average
numbers should be less than 1 ms.
Additional RAC statistics are then organized in the following sections:
The Global Enqueue Statistics section contains data extracted from
V$GES_STATISTICS.
The Global CR Served Stats section contains data from V$CR_BLOCK_SERVER.

The Global CURRENT Served Stats section contains data from
V$CURRENT_BLOCK_SERVER.
The Global Cache Transfer Stats section contains data from
V$INSTANCE_CACHE_TRANSFER.
The Segment Statistics section also includes the GC Buffer Busy Waits, CR Blocks Received,
and CUR Blocks Received information for relevant segments.
s a
) ha
Note: For more information about wait events and statistics, refer to Oracle Database
Reference. m
o
c e
y s
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Active Session History Reports for RAC
Active Session History (ASH) report statistics provide

details about the RAC Database session activity.

The database records
information about active
sessions for all active RAC
instances.
s a
)ha
m
co
s
u n isy uide
Two ASH report sections b r nt G
specific to Oracle RAC are n t o@ tude
e S
Top Cluster Events and scim this
Top Remote Instance. n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Activeto n E History
Session n s fe(ASH) is an integral part of the Oracle Database self-management
v e r nand
framework - trisauseful for diagnosing performance problems in Oracle RAC environments.
no statistics provide details about Oracle Database session activity. Oracle Database
EASH report
records information about active sessions for all active Oracle RAC instances and stores this
data in the System Global Area (SGA). Any session that is connected to the database and
using CPU is considered an active session. The exception to this is sessions that are waiting
for an event that belongs to the idle wait class.
ASH reports present a manageable set of data by capturing only information about active
sessions. The amount of the data is directly related to the work being performed, rather than
the number of sessions allowed on the system. ASH statistics that are gathered over a
specified duration can be put into ASH reports.
Each ASH report is divided into multiple sections to help you identify short-lived performance
problems that do not appear in the ADDM analysis. Two ASH report sections that are specific
to Oracle RAC are Top Cluster Events and Top Remote Instance.

Top Cluster Events
The ASH report Top Cluster Events section is part of the Top Events report that is specific to
Oracle RAC. The Top Cluster Events report lists events that account for the highest
percentage of session activity in the cluster wait class event along with the instance number
of the affected instances. You can use this information to identify which events and instances
caused a high percentage of cluster wait events.
Top Remote Instance

The ASH report Top Remote Instance section is part of the Top Load Profile report that is
specific to Oracle RAC. The Top Remote Instance report shows cluster wait events along with
the instance numbers of the instances that accounted for the highest percentages of session
activity. You can use this information to identify the instance that caused the extended cluster
wait period.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Automatic Database Diagnostic Monitor for RAC
ADDM can perform analysis:
For the entire cluster
For a specific database instance

Database ADDM
For a subset of database instances
Self-diagnostic engine
s a
)ha
Instance ADDM m
co
s
u n isy uide
b r nt G
t o @ tude
AWR
ime is S n
a s c e th
Inst1
o n n o us Instn
r t t
( e ve ense
o n n e2012,
Copyright
v e rt rab
n fe
E nsDatabase
r o
Usingtthe Automatic
t r a Diagnostic Monitor (ADDM), you can analyze the information
ve obynAWR
Ecollected - for possible performance problems with your Oracle database. ADDM
presentsn performance data from a clusterwide perspective, thus enabling you to analyze
performance on a global basis. In an Oracle RAC environment, ADDM can analyze
performance using data collected from all instances and present it at different levels of
granularity, including:
Analysis for the entire cluster
Analysis for a specific database instance
Analysis for a subset of database instances
To perform these analyses, you can run the ADDM Advisor in Database ADDM for RAC mode
to perform an analysis of the entire cluster, in Local ADDM mode to analyze the performance
of an individual instance, or in Partial ADDM mode to analyze a subset of instances.
Database ADDM for RAC is not just a report of reports but has independent analysis that is
appropriate for RAC. You activate ADDM analysis using the advisor framework through
Advisor Central in Oracle Enterprise Manager, or through the DBMS_ADVISOR and
DBMS_ADDM PL/SQL packages.
Note: Database ADDM report is generated on AWR snapshot coordinator.

Automatic Database Diagnostic Monitor for RAC
Identifies the most critical performance problems for the

entire RAC cluster database

Runs automatically when taking AWR snapshots
Performs database-wide analysis of:
Global resources (for example I/O and global locks)
High-load SQL and hot blocks
Global cache interconnect traffic s a
) ha
Network latency issues m
c o
Skew in instance response times ys e
is id
u
Is used by DBAs to analyze cluster performance
r n Gu
@ b ent
Does not require investigation ofnnto tuto
reports d spot common
e
m this S
problems sci n a use

e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
In Oraclen E ns11g,
Database
fe you can create a period analysis mode for ADDM that analyzes the
e o tra
rt nperformance
throughput
v
Edatabase o - for an entire cluster. When the advisor runs in this mode, it is called
n ADDM. You can run the advisor for a single instance, which is called instance
ADDM.
Database ADDM has access to AWR data generated by all instances, thereby making the
analysis of global resources more accurate. Both database and instance ADDM run on
continuous time periods that can contain instance startup and shutdown. In the case of
database ADDM, there may be several instances that are shut down or started during the
analysis period. However, you must maintain the same database version throughout the
entire time period.
Database ADDM runs automatically after each snapshot is taken. You can also perform
analysis on a subset of instances in the cluster. This is called partial analysis ADDM.
I/O capacity finding (the I/O system is overused) is a global finding because it concerns a
global resource affecting multiple instances. A local finding concerns a local resource or issue
that affects a single instance. For example, a CPU-bound instance results in a local finding
about the CPU. Although ADDM can be used during application development to test changes
to either the application, the database system, or the hosting machines, database ADDM is
targeted at DBAs.

What Does ADDM Diagnose for RAC?
Latency problems in interconnect

Congestion (identifying top instances affecting the entire

cluster)
Contention (buffer busy, top objects, and so on)
Top consumers of multiblock requests
Lost blocks
Reports information about interconnect devices; warns has
a
about using PUBLIC interfaces m )
o
c e
Reports throughput of devices, and how much y s
u n is it uisidused
of
br PQ)
by Oracle and for what purpose (GC, locks, nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E are:nsfe
e t o
Data sources
rWait tra (especially Cluster class and buffer busy)
v n -
events
E

no
Active Session History (ASH) reports
Instance cache transfer data
Interconnect statistics (throughput, usage by component, pings)
ADDM analyzes the effects of RAC for both the entire database (DATABASE analysis mode)
and for each instance (INSTANCE analysis mode).

EM Support for ADDM for RAC
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E 11g s e
fEnterprise
Oracle
r t o
Database
t r a n Manager displays the ADDM analysis on the Cluster
ve oHome
EDatabase n- page.
n
On the Automatic Database Diagnostic Monitor (ADDM) page, the Database Activity chart
(not shown here) plots the database activity during the ADDM analysis period. Database
activity types are defined in the legend based on its corresponding color in the chart. Each
icon below the chart represents a different ADDM task, which in turn corresponds to a pair of
individual Oracle Database snapshots saved in the Workload Repository.
In the ADDM Performance Analysis section, the ADDM findings are listed in descending
order, from highest impact to least impact. For each finding, the Affected Instances column
displays the number (m of n) of instances affected. Drilling down further on the findings takes
you to the Performance Findings Detail page. The Informational Findings section lists the
areas that do not have a performance impact and are for informational purpose only.
The Affected Instances chart shows how much each instance is impacted by these findings.
The display indicates the percentage impact for each instance.

Quiz
Although there are specific tuning areas for RAC, such as

instance recovery and interconnect traffic, you get most

benefits by tuning your system like a single-instance system.
a. True
b. False
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o a
rt n-tra
v
E no

Quiz
Which of the following RAC tuning tips are correct?

a. Application tuning is often the most beneficial!

b. Reduce long full-table scans in OLTP systems.
c. Eliminate sequence caches.
d. Use partitioning to reduce inter-instance traffic.
e. Configure the interconnects properly.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E d, ensfe
Answer:
e o a, b,
rt n-tra
v
E no c is incorrect.
Statement

Summary

Determine RAC-specific tuning components

Determine RAC-specific wait events, global enqueues, and
system statistics
Implement the most common RAC tuning tips
Use the Cluster Database Performance pages
Use the Automatic Workload Repository (AWR) in RAC has
a
m )
Use Automatic Database Diagnostic Monitor (ADDM) o in
y s c e
RAC is id
r u n Gu
@ b ent
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Practice 7 Overview
This practice covers manually discovering performance issues

by using the EM performance pages as well as ADDM.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Managing High Availability of Services
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives

Configure and manage services in a RAC environment

Use services with client applications
Use services with the Database Resource Manager
Use services with the Scheduler
Configure services aggregation and tracing
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Oracle Services
To manage workloads or a group of applications, you can

define services for a particular application or a subset of an

applications operations.
You can also group work by type under services.
For example OLTP users can use one service while batch
processing can use another to connect to the database.
Users who share a service should have the same service- s a
level requirements. ) ha
m
Use srvctl or Enterprise Manager to manageysservices, co e
not DBMS_SERVICE. u n is uid
r G b ent
@
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eworkloads
n s feor a group of applications, you can define services that you assign to a
e o
To manage
tra or to a subset of an applications operations. You can also group work
rt application
v
particular n -
E no
by type under services. For example, online users can use one service, while batch
processing can use another, and reporting can use yet another service to connect to the
database.
It is recommended that all users who share a service have the same service-level
requirements. You can define specific characteristics for services and each service can be a
separate unit of work. There are many options that you can take advantage of when using
services. Although you do not have to implement these options, using them helps optimize
application performance. You can define services for both policy-managed and administrator-
managed databases.
Do not use DBMS_SERVICE with cluster-managed services. When Oracle Clusterware starts
a service, it updates the database with the attributes stored in the CRS resource. If you use
DBMS_SERVICE to modify the service and do not update the CRS resource, the next time
CRS resource is started, it will override the database attributes set by DBMS_SERVICE.

Services for Policy-
and Administrator-Managed Databases
You can define services for both policy-managed and
administrator-managed databases.
Services for a policy-managed database are defined to a
server pool where the database is running.
Services for policy-managed databases can be defined as:
UNIFORM (running on all instances in the server pool)
SINGLETON (running on only one instance in the pool)
For singleton services, RAC chooses on which instance in theha
sa

)
server pool the service is active. om c
Services for an administrator-managed database i s ys define
i d e
r u n Gu
which instances normally support that service.
b nt
@ e
These are known as the PREFERRED
e tud
nto Sinstances.
c
Instances defined to support
s e t his if the preferred
ima service
a us
nn as AVAILABLE
instance fails are known
r to e to
instances.
e e
v ens
(
n e2012,
o n
rt rab l
v e
n E nthat s feall users who share a service have the same service-level
e o
It is recommended
rt n-tYou ra can define specific characteristics for services and each service can be a
v
requirements.
Eseparate nounit of work. There are many options that you can take advantage of when using
services. Although you do not have to implement these options, they help optimize application
performance. You can define services for both policy-managed and administrator-managed
databases.
Policy-managed database: When you define services for a policy-managed database,
you define the service to a server pool where the database is running. You can define
the service as either uniform (running on all instances in the pool) or singleton (running
on only one instance in the pool). For singleton services, RAC chooses on which
instance in the server pool the service is active. If that instance fails, then the service
fails over to another instance in the pool. A service can run in only one server pool.
Administrator-managed database: When you define a service for an administrator-
managed database, you define which instances support that service. These are known
as the PREFERRED instances. You can also define other instances to support a service
if the services preferred instance fails. These are known as AVAILABLE instances.
Note: Failback is not implemented by default because it is a disruptive operation. If you
require failback, you can always try to implement it as a callout that relocates the service.

Default Service Connections
Application services:
Limit of 115 services per database

Internal services:
SYS$BACKGROUND
SYS$USERS
Cannot be deleted or changed
A special Oracle database service is created by default foras a
the Oracle RAC database. m )h
s co in
This default service is always available on all sinstances
u n i y uide
an Oracle RAC environment. r tG
@ b en
e nt o
S tud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nstypes fe of services: application services and internal services. Application
e o
Theretare two broad
tra functional maps to workloads. Sessions doing work for a common
r aren-mainly
v
services
nofunction are grouped together. For Oracle E-Business Suite, AP, AR, GL, MFG, WIP,
Ebusiness
and so on create a functional division of work within the database and can be categorized as
services.
The RDBMS also supports two internal services. SYS$BACKGROUND is used by the
background processes only. SYS$USERS is the default service for user sessions that are not
associated with any application service. Both internal services support all the workload
management features and neither one can be stopped or disabled. A special Oracle database
service is created by default for your Oracle RAC database. This default service is always
available on all instances in an Oracle RAC environment, unless an instance is in restricted
mode. You cannot alter this service or its properties. There is a limitation of 115 application
services per database that you can create. Also, a service name is restricted to 64 characters.
Note: Shadow services are also included in the application service category. For more
information about shadow services, see the lesson titled High Availability of Connections. In
addition, a service is also created for each Advanced Queue created. However, these types of
services are not managed by Oracle Clusterware. Using service names to access a queue
provides location transparency for the queue within a RAC database.

Creating Service with Enterprise Manager
Administration-Managed
Policy-Managed
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
From yourn E s fe home page, click the Availability tab, and then click Cluster
ClusternDatabase
e o tra Services. On the Cluster Managed Database Services page, click Create
rt Database
Managed
v
EService.no n -
Use the Create Service page to configure a new service in which you do the following:
Select the desired service policy for each instance configured for the cluster database.
Select the desired service properties.
If your database is administration managed, the High Availability Configuration section allows
you to configure preferred and available servers. If your database employs policy-managed
administration, you can configure the service cardinality to be UNIFORM or SINGLETON and
assign the service to a server pool.
You can also define the management policy for a service. You can choose either an
automatic or a manual management policy.
Automatic: The service always starts when the database starts.
Manual: Requires that the service be started manually. Prior to Oracle RAC 11g
Release 2, all services worked as though they were defined with a manual management
policy.
Note: Enterprise Manager now generates the corresponding entries in your tnsnames.ora
files for your services. Just click the Update local naming parameter (tnsnames.ora) file
check box when creating the service.

Creating Services with SRVCTL
To create a service called GL with preferred instance

RAC02 and an available instance RAC01:
$ srvctl add service d PROD1 s GL -r RAC02 -a RAC01
To create a service called AP with preferred instance

RAC01 and an available instance RAC02:
$ srvctl add service d PROD1 s AP r RAC01 -a RAC02
To create a SINGLETON service called BATCH using server s a

) ha
pool SP1 and a UNIFORM service called ERP using server
m
o
c e
pool SP2: s y s d
u n i u i
$ srvctl add service -d PROD2 -s BATCH -g
br SP1e\nt G
@
nto Stud
-c singleton -y manual
im e is SP2 \
$ srvctl add service -d PROD2 s c -s ERP
e t h-g
-c UNIFORM -y manual nn
a us
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E innthe s feslide, assume a two-node, administration-managed database called
For the
e oexample
rt withn-antrainstance named RAC01 on one node and an instance called RAC02 on the
v
PROD1
no services are created, AP and GL, to be managed by Oracle Clusterware. The AP
Eother. Two
service is defined with a preferred instance of RAC01 and an available instance of RAC02.
If RAC01 dies, the AP service member on RAC01 is restored automatically on RAC02. A
similar scenario holds true for the GL service.
Note that it is possible to assign more than one instance with both the -r and -a options.
However, -r is mandatory but -a is optional.
Next, assume a policy-managed cluster database called PROD2. Two services are created, a
SINGELTON service called BATCH and a UNIFORM service called ERP. SINGLETON services
run on one of the active servers and UNIFORM services run on all active servers of the server
pool. The characteristics of the server pool determine how resources are allocated to the
service.
Note: When services are created with srvctl, tnsnames.ora is not updated and the
service is not started.

Managing Services with Enterprise Manager
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s feManager to manage services within a GUI framework. The
EEnterprise
You can
e o use
raslide shows the main page for administering services within RAC. It shows
rt nin-tthe
v
screenshot
nobasic status information about a defined service.
Eyou some
To access this page, click the Cluster Managed Database Services link on the Cluster
Database Availability page.
You can perform simple service management such as enabling, disabling, starting, stopping,
and relocating services. All possible operations are shown in the slide.
If you choose to start a service on the Cluster Managed Database Services page, then EM
attempts to start the service on every preferred instance. Stopping the service stops it on all
instances that it is currently running.
To relocate a service, select the service that you want to administer, select the Manage option
from the Actions drop-down list, and then click Go.
Note: On the Cluster Managed Database Services page, you can test the connection for a
service.

Managing Services with EM
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Cluster
n s feManaged Database Service page for an individual service, you must
e o
To access the
trafrom the Cluster Managed Database Services page, select the Manage
rta service
v
select n -
no the Actions drop-down list, and then click Go.
Eoption from
This is the Cluster Managed Database Service page for an individual service. It offers you the
same functionality as the previous page, except that actions performed here apply to specific
instances of a service.
This page also offers you the added functionality of relocating a service to an available
instance. Relocating a service from one instance to another stops the service on the first
instance and then starts it on the second.

Managing Services with srvctl
Start a named service on all configured instances:

$ srvctl start service d orcl s AP
Stop a service:
$ srvctl stop service d orcl s AP I orcl3,orcl4
Disable a service at a named instance:

s a
$ srvctl disable service d orcl s AP i orcl4
)ha
m
co
Set an available instance as a preferred instance: s
n i sy uide
$ srvctl modify service d orcl s AP -irorcl5 r
b u t G
de n
Relocate a service from one instance n t o@to another:t u
c i me this S
$ srvctl relocate servicend
as orcl u s es AP -i orcl5 t orcl4
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
The sliden E nsfesome management tasks with services by using SRVCTL.
demonstrates
e o
rt that traAP service has been created with four preferred instances: orcl1, orcl2,
v
Assume n -an
o orcl4. An available instance, orcl5, has also been defined for AP.
Eorcl3,nand
In the first example, the AP service is started on all instances. If any of the preferred or
available instances that support AP are not running but are enabled, then they are started.
The stop command stops the AP service on instances orcl3 and orcl4. The instances
themselves are not shut down, but remain running possibly supporting other services. The AP
service continues to run on orcl1 and orcl2. The intention might have been to perform
maintenance on orcl4, and so the AP service was disabled on that instance to prevent
automatic restart of the service on that instance. The OCR records the fact that AP is disabled
for orcl4. Thus, Oracle Clusterware will not run AP on orcl4 until the service is enabled.
The next command in the slide changes orcl5 from being an available instance to a
preferred one. This is beneficial if the intent is to always have four instances run the service
because orcl4 was previously disabled. The last example relocates the AP service from
instance orcl5 to orcl4. Do not perform other service operations while the online service
modification is in progress.

Using Services with Client Applications
ERP=(DESCRIPTION= ## Using the SCAN ##

(LOAD_BALANCE=on)
(ADDRESS=(PROTOCOL=TCP)(HOST=cluster01-scan)(PORT=1521))
(CONNECT_DATA=(SERVICE_NAME=ERP)))
ERP=(DESCRIPTION= ## Using VIPs ##

(LOAD_BALANCE=on)
(ADDRESS=(PROTOCOL=TCP)(HOST=node1-vip)(PORT=1521))
s a
)ha
m
co
s
url="jdbc:oracle:oci:@ERP" ## Thick JDBC ##
u n isy uide
b r nt G
url="jdbc:oracle:thin:@(DESCRIPTION= to@## Thin
n t u deJDBC ##
(LOAD_BALANCE=on)
c i me this S
n as use
(ADDRESS=(PROTOCOL=TCP)(HOST=cluster01-scan)(PORT=1521)))
ton e to
(CONNECT_DATA=(SERVICE_NAME=ERP)))"
e r
v ens
( e
o n n e2012,
Copyright
v e rt rab
n E ninsthe feslide shows the TNS connect descriptor that can be used to access
e o
The first example
rt service.traIt uses the clusters Single Client Access Name (SCAN). The SCAN provides
the
v ERP n -
o to the clients connecting to Oracle RAC that does not change throughout the
Ea singlenname
life of the cluster, even if you add or remove nodes from the cluster. Clients connecting with
SCAN can use a simple connection string, such as a thin JDBC URL or EZConnect, and still
achieve the load balancing and client connection failover. The second example uses virtual IP
addresses as in previous versions of the Oracle Database.
The third example shows the thick JDBC connection description using the previously defined
TNS connect descriptor.
The third example shows the thin JDBC connection description using the same TNS connect
descriptor as the first example.
Note: The LOAD_BALANCE=ON clause is used by Oracle Net to randomize its progress
through the protocol addresses of the connect descriptor. This feature is called client
connection load balancing.

Services and Connection Load Balancing
The two load balancing methods that you can implement are:
Client-side load balancing: Balances the connection
requests across the listeners

Server-side load balancing: The listener directs a
connection request to the best instance currently providing
the service by using the load balancing advisory (LBA).
FAN, Fast Connection Failover, and LBA depend on a
connection load balancing configuration that includes a
ha s
setting the connection load balancing goal for the service.
)
o m
sc
The load balancing goal for the service can be either: sy e
n i
LONG: For applications having long-lived connections.
u u idThis is

br sessions.n t G
typical for connection pools and SQL*Forms
@ e
o tud connections
ent short-lived
SHORT: For applications that have S
s c im thi
n s
aservice_name
s e
srvctl modify service -s
n o u -j LONG|SHORT
v e rto se t
(e 2012, c n and/or its affiliates. All rights reserved.
eOracle
n n
Copyright l i
e r t o
a b le
n E v
s f er the ability to balance client connections across the instances in
Oracleo
r Net
t RAC Services
t r a n provides
an
v e
Oracle
o
Eserver-side.n - configuration. You can implement two types of load balancing: client-side and
n Client-side load balancing balances the connection requests across the listeners.
With server-side load balancing, the listener directs a connection request to the best instance
currently providing the service by using the load balancing advisory. In a RAC database,
client connections should use both types of connection load balancing.
FAN, Fast Connection Failover, and the load balancing advisory depend on an accurate
connection load balancing configuration that includes setting the connection load balancing
goal for the service. You can use a goal of either LONG or SHORT for connection load
balancing. These goals have the following characteristics:
LONG: Use the LONG load balancing method for applications that have long-lived
connections. This is typical for connection pools and SQL*Forms sessions. LONG is the
default connection load balancing goal. The following is an example of modifying a
service, POSTMAN, with the srvctl utility to define the connection load balancing goal
for long-lived sessions: srvctl modify service -s POSTMAN -j LONG
SHORT: Use the SHORT connection load balancing method for applications that have
short-lived connections. The following example modifies the ORDER service, using
srvctl to set the goal to SHORT: srvctl modify service -s ORDER -j SHORT

Services and Transparent Application Failover
Services simplify the deployment of Transparent

Application Failover (TAF).
You can define a TAF policy for a service and all

connections using this service will automatically have TAF
enabled.
The TAF setting on a service can be NONE, BASIC, or
PRECONNECT and overrides any TAF setting in the client
connection definition. s a
ha
)can
To define a TAF policy for a service, the srvctl utility
o m
y s c e
be used as follows: is id
r u n Gu
srvctl modify service -s gl.example.com b-q TRUE
e n t -P BASIC
@
-e SELECT -z 180 -w 5 -j LONG
e nto Stud
Where -z is the number of retries, -w is the s c im between
delay
e t his retry attempts and -j is the
connection load balancing goal
n na us
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe establishes a connection to an instance, the connection remains
E Net nServices
e o
WhentOracle
runtil the ra closes the connection, the instance is shut down, or a failure occurs. If
tclient
v
open n -
no TAF for the connection, then Oracle Database moves the session to a surviving
Eyou configure
instance when an outage occurs.
TAF can restart a query after failover has completed but for other types of transactions, such
as INSERT, UPDATE, or DELETE, the application must roll back the failed transaction and
resubmit the transaction. You must re-execute any session customizations, in other words,
ALTER SESSION statements, after failover has occurred. However, with TAF, a connection is
not moved during normal processing, even if the workload changes over time.
Services simplify the deployment of TAF. You can define a TAF policy for a service, and all
connections using this service will automatically have TAF enabled. This does not require any
client-side changes. The TAF setting on a service overrides any TAF setting in the client
connection definition. To define a TAF policy for a service, use the srvctl utility as in the
following example:
srvctl modify service -s gl.example.com -q TRUE -P BASIC -e
SELECT -z 180 -w 5 -j LONG
Note: TAF applies only to an admin-managed database and not to policy-managed
databases.

Using Services with the Resource Manager
Consumer groups are automatically assigned to sessions

based on session services.

Work is prioritized by service inside one instance.
Instance resources s a
AP
)ha
m
co
s
Connections AP
u n isy uide
75%
b r nt G
t o @ tude
n S 25%
BATCH
c i meBATCH h i s
n as use t
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E Resource
n s fe Manager (also called Resource Manager) enables you to identify
e o
The Database
ra
t using-tservices.
rby
v
work
o n
Ebindingnservices It manages the relative priority of services within an instance by
directly to consumer groups. When a client connects by using a service, the
consumer group is assigned transparently at connect time. This enables the Resource
Manager to manage the work requests by service in the order of their importance.
For example, you define the AP and BATCH services to run on the same instance, and assign
AP to a high-priority consumer group and BATCH to a low-priority consumer group. Sessions
that connect to the database with the AP service specified in their TNS connect descriptor get
priority over those that connect to the BATCH service.
This offers benefits in managing workloads because priority is given to business functions
rather than the sessions that support those business functions.

Services and Resource Manager with EM
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E ns(EM) fe presents a GUI through the Consumer Group Mapping page to
e o
Enterprise Manager
rt n-tmap ra sessions to consumer groups. You can access this page by clicking the
v
automatically
no Group Mappings link on the Server page.
EConsumer
Using the General tabbed page of the Consumer Group Mapping page, you can set up a
mapping of sessions connecting with a service name to consumer groups as illustrated in the
right half of the slide.
With the ability to map sessions to consumer groups by service, module, and action, you have
greater flexibility when it comes to managing the performance of different application
workloads.
Using the Priorities tabbed page of the Consumer Group Mapping page, you can change
priorities for the mappings that you set up on the General tabbed page. The mapping options
correspond to columns in V$SESSION. When multiple mapping columns have values, the
priorities you set determine the precedence for assigning sessions to consumer groups.
Note: You can also map a service to a consumer group directly from the Create Service page
as shown in the left half of the slide.

Using Services with the Scheduler
Services are associated with Scheduler job classes.

Scheduler jobs have service affinity:

High availability
Load balancing
HOT_BATCH_SERV HOT_BATCH_SERV LOW_BATCH_SERV
s a
Job coordinator Job coordinator Job coordinator
) ha
Job slaves Job slaves Job slaves c o m
i s ys ide
Database r u n Gu
@ b ent
Job table
e nto Stud
his
Job1 HOT_BATCH_CLASS HOT_BATCH_SERV
Job2 HOT_BATCH_CLASS
s c imHOT_BATCH_SERV
t
na us
Job3 LOW_BATCH_CLASS
e
LOW_BATCH_SERV
r n
to e to
e e
v ens
(
n e2012,
o
rt rabn l
v e
n n s fe
E environments,
Just as
r t oin other
t r a the Scheduler in a RAC environment uses one job table for
ve database
Eeach - and one job coordinator (CJQ0 process) for each instance. The job
non communicate with each other to keep information current.
coordinators
The Scheduler can use the services and the benefits they offer in a RAC environment. The
service that a specific job class uses is defined when the job class is created. During
execution, jobs are assigned to job classes and job classes run within services. Using
services with job classes ensures that the work of the Scheduler is identified for workload
management and performance tuning. For example, jobs inherit server-generated alerts and
performance thresholds for the service they run under.
For high availability, the Scheduler offers service affinity instead of instance affinity. Jobs are
not scheduled to run on any specific instance. They are scheduled to run under a service. So,
if an instance dies, the job can still run on any other instance in the cluster that offers the
service.
Note: By specifying the service where you want the jobs to run, the job coordinators balance
the load on your system for better performance.

Services and the Scheduler with EM
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe under a specific service, click the Job Classes link in the Database
Ea jobntosrun
e o
To configure
tra of the Server page. This opens the Scheduler Job Classes page. On the
rt nsection
v
Scheduler -
no Job Classes page, you can see services assigned to job classes.
EScheduler
When you click the Create button on the Scheduler Job Classes page, the Create Job Class
page is displayed. On this page, you can enter details of a new job class, including which
service it must run under.
Note: Similarly, you can map a service to a job class on the Create Service page as shown at
the bottom of the slide.

Services and the Scheduler with EM
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n e
E classnsisfset
r t o
After your job
t r a up with the service that you want it to run under, you can create the
veTo create
Ejob. - the job, click the Jobs link on the Server page. The Scheduler Jobs page
appears, noonnwhich you can click the Create button to create a new job. When you click the
Create button, the Create Job page is displayed. This page has different tabs: General,
Schedule, and Options. Use the General tabbed page to assign your job to a job class.
Use the Options page (displayed in the slide) to set the Instance Stickiness attribute for your
job. Basically, this attribute causes the job to be load balanced across the instances for which
the service of the job is running. The job can run only on one instance. If the Instance
Stickiness value is set to TRUE, which is the default value, the Scheduler runs the job on the
instance where the service is offered with the lightest load. If Instance Stickiness is set to
FALSE, then the job is run on the first available instance where the service is offered.
Note: It is possible to set job attributes, such as INSTANCE_STICKINESS, by using the
SET_ATTRIBUTE procedure of the DBMS_SCHEDULER PL/SQL package.

Using Distributed Transactions with RAC
An XA transaction can span RAC instances, allowing any

application that uses XA to take full advantage of the

Oracle RAC environment.
Tightly coupled XA transactions no longer require the
special type of singleton services (DTP).
XA transactions are transparently supported on Oracle
RAC databases with any type of services configuration. a
However, DTP services will improve performance for many ha s
m )
distributed transaction scenarios. o
c e
y s
DTP services allow you to direct all branches
u n isof a uid
distributed transaction to a single instance b
r in the n t G
cluster.
o @ d e
To load balance, it is better to have
e nt several
S tu groups of
smaller application servers s c im each
with e t hisgroup directing its
a us
nservice.
to n
transactions to a single to
e v er nse
n ( 2012,
r n
to able
e
Ev ncan r
An XAto n
transaction s fespan Oracle RAC instances by default, allowing any application that
r t r a
ve XA otontake
Euses - full advantage of the Oracle RAC environment to enhance the availability and
n of the application. This is controlled through the GLOBAL_TXN_PROCESSES
scalability
initialization parameter, which is set to 1 by default. This parameter specifies the initial
number of GTXn background processes for each Oracle RAC instance. Keep this parameter
at its default value clusterwide to allow distributed transactions to span multiple Oracle RAC
instances. This allows the units of work performed across these Oracle RAC instances to
share resources and act as a single transaction (that is, the units of work are tightly coupled).
It also allows 2PC requests to be sent to any node in the cluster. Tightly coupled XA
transactions no longer require the special type of singleton services (that is, Oracle
Distributed Transaction Processing [DTP] services) to be deployed on Oracle RAC database.
XA transactions are transparently supported on Oracle RAC databases with any type of
services configuration.
To provide improved application performance with distributed transaction processing in
Oracle RAC, you may want to take advantage of the specialized service referred to as a DTP
Service. Using DTP services, you can direct all branches of a distributed transaction to a
single instance in the cluster. To load balance across the cluster, it is better to have several
groups of smaller application servers with each group directing its transactions to a single
service, or set of services, than to have one or two larger application servers.

Distributed Transactions and Services
To leverage all instances, create one or more DTP services for
each instance that hosts distributed transactions.
Use EM or srvctl to create singleton services XA1, XA2,

and XA3 for database CRM, enabling DTP for the service.
EM is used to create service XA1 as follows:
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe of distributed transactions, you can use services to manage
Ethe performance
e t o
To enhance
tra By defining the DTP property of a service, the service is guaranteed to
renvironments.
v
DTP n -
o instance at a time in an Oracle RAC database. All global distributed transactions
Erun on none
performed through the DTP service are ensured to have their tightly coupled branches
running on a single Oracle RAC instance. This has the following benefits:
The changes are available locally within one Oracle RAC instance when tightly coupled
branches need information about changes made by each other.
Relocation and failover of services are fully supported for DTP.
By using more DTP services than there are Oracle RAC instances, Oracle Database
can balance the load by services across all the Oracle RAC database instances.
To leverage all the instances in a cluster, create one or more DTP services for each Oracle
RAC instance that hosts distributed transactions. Choose one DTP service for one distributed
transaction. Choose different DTP services for different distributed transactions to balance the
workload among the Oracle RAC database instances.
Because all the branches of a distributed transaction are on one instance, you can leverage
all the instances to balance the load of many DTP transactions through multiple singleton
services, thereby maximizing application throughput.

An external transaction manager, such as OraMTS, coordinates DTP/XA transactions.
However, an internal Oracle transaction manager coordinates distributed SQL transactions.
Both DTP/XA and distributed SQL transactions must use the DTP service in Oracle RAC.
For services that you are going to use for distributed transaction processing, create the
service by using Oracle Enterprise Manager or srvctl, and define only one instance as the
preferred instance. You can have as many AVAILABLE instances as you want. For example,
the following srvctl command creates a singleton service for database CRM,
xa1.example.com, whose preferred instance is orcl1:

srvctl add service -d CRM -s xa1.example.com -r orcl1 -a orcl2,
orcl3
Then mark the service for distributed transaction processing by setting the DTP parameter to
TRUE; the default is FALSE.
srvctl modify service -d CRM -s xa1.example.com -x TRUE
Conversely, you can create the service and mark it for distributed transaction processing with
s a
a single command if desired:
)h a
srvctl add service -d CRM -s xa1.example.com -r orcl1 -a orcl2, m
co
s
orcl3 x TRUE
u n isy uide
Oracle Enterprise Manager enables you to set this parameter on
b n t G Managed
r the Cluster
Database Services: Create Service or Modify Service page.
t o @ tude
n
e then
If, for example, orcl1 (that provides service XA1) fails, S singleton service that it
provided fails over to another instance, suchsasc im h
orcl2 tor
s the
iorcl3. If services migrate to other
a us e
n RACo database,
instances after the cold-start of the Oracle then you might have to force the
t o n
relocation of the service to evenlyr re-balancetthe load on all the available hardware. Use data
from the GV$ACTIVE_SERVICES e v e viewntosedetermine whether to do this.
n ( lice
r t o n le
e a b
n Ev nsfer
e r to -tra
Ev non

Service Thresholds and Alerts
Service-level thresholds enable you to compare achieved
service levels against accepted minimum required levels.
You can explicitly specify two performance thresholds for

each service:
SERVICE_ELAPSED_TIME: The response time for calls
SERVICE_CPU_TIME: The CPU time for calls
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Ethresholds
n s fe enable you to compare achieved service levels against accepted
e o
Service-level
tra levels. This provides accountability for the delivery or the failure to deliver
rt required
v
minimum n -
noservice level. The end goal is a predictable system that achieves service levels.
Ean agreed
There is no requirement to perform as fast as possible with minimum resource consumption;
the requirement is to meet the quality of service.
You can explicitly specify two performance thresholds for each service: the response time for
calls, or SERVICE_ELAPSED_TIME, and the CPU time for calls, or SERVICE_CPU_TIME. The
response time goal indicates that the elapsed time should not exceed a certain value, and the
response time represents wall clock time. Response time is a fundamental measure that
reflects all delays and faults that might be blocking the call from running on behalf of the user.
Response time can also indicate differences in node power across the nodes of an Oracle
RAC database.
The service time and CPU time are calculated as the moving average of the elapsed, server-
side call time. The AWR monitors the service time and CPU time and publishes AWR alerts
when the performance exceeds the thresholds. You can then respond to these alerts by
changing the priority of a job, stopping overloaded processes, or by relocating, expanding,
shrinking, starting, or stopping a service. This permits you to maintain service availability
despite changes in demand.

Services and Thresholds Alerts: Example
EXECUTE DBMS_SERVER_ALERT.SET_THRESHOLD(
METRICS_ID => DBMS_SERVER_ALERT.ELAPSED_TIME_PER_CALL
, warning_operator => DBMS_SERVER_ALERT.OPERATOR_GE
, warning_value => '500000'
, critical_operator => DBMS_SERVER_ALERT.OPERATOR_GE
, critical_value => '750000'
, observation_period => 30
, consecutive_occurrences => 5
s a
, instance_name => NULL
) ha
o
, object_type => DBMS_SERVER_ALERT.OBJECT_TYPE_SERVICE
c m
, object_name => 'servall'); ys e is uid
r u n G
b n t
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
To check n E nsfe for the servall service, use the AWR report. You should record
the thresholds
e o
rt fromnthe a over several successive intervals during which time the system is
trreport
v
output -
E no
running optimally. For example, assume that for an email server, the AWR report runs each
Monday during the peak usage times of 10:00 AM to 2:00 PM. The AWR report would contain
the response time, or DB time, and the CPU consumption time, or CPU time, for calls for each
service. The AWR report would also provide a breakdown of the work done and the wait times
that are contributing to the response times.
Using DBMS_SERVER_ALERT, set a warning threshold for the servall service at 0.5
seconds and a critical threshold for the payroll service at 0.75 seconds. You must set these
thresholds at all instances within an Oracle RAC database. The parameter instance_name
can be set to a NULL value to indicate database-wide alerts. You can schedule actions using
Enterprise Manager jobs for alerts, or you can schedule actions to occur programmatically
when the alert is received. In this example, thresholds are added for the servall service and
set as shown in the slide.
Verify the threshold configuration by using the following SELECT statement:
SELECT metrics_name, instance_name, warning_value,
critical_value, observation_period FROM dba_thresholds;

Service Aggregation and Tracing
Statistics are always aggregated by service to measure

workloads for performance tuning.

Statistics can be aggregated at finer levels:
MODULE
ACTION
Combination of SERVICE_NAME, MODULE, ACTION
Tracing can be done at various levels: a
SERVICE_NAME ha s
m )
MODULE o
c e
y s
ACTION
u n is uid
b
r ACTIONn t G
Combination of SERVICE_NAME, MODULE,@ e
This is useful for tuning systemsm
to
St sessions.
enuseisshared
that
ud
i

s
na us
ce th
r n
to e to
e e
v ens
(
n e2012,
o n
rt rab l
v e
n fe
E nsstatistics
r t o
By default, important
t r a and wait events are collected for the work attributed to every
ve An
Eservice. -
application can further qualify a service by MODULE and ACTION names to identify
non transactions within the service. This enables you to locate exactly the poorly
the important
performing transactions for categorized workloads. This is especially important when
monitoring performance in systems by using connection pools or transaction processing
monitors. For these systems, the sessions are shared, which makes accountability difficult.
SERVICE_NAME, MODULE, and ACTION are actual columns in V$SESSION. SERVICE_NAME
is set automatically at login time for the user. MODULE and ACTION names are set by the
application by using the DBMS_APPLICATION_INFO PL/SQL package or special OCI calls.
MODULE should be set to a user-recognizable name for the program that is currently
executing. Likewise, ACTION should be set to a specific action or task that a user is
performing within a module (for example, entering a new customer).
Another aspect of this workload aggregation is tracing by service. The traditional method of
tracing each session produces trace files with SQL commands that can span workloads. This
results in a hit-or-miss approach to diagnose problematic SQL. With the criteria that you
provide (SERVICE_NAME, MODULE, or ACTION), specific trace information is captured in a set
of trace files and combined into a single output trace file. This enables you to produce trace
files that contain SQL that is relevant to a specific workload being done.

Top Services Performance Page
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfepage, you can access the Top Consumers page by clicking the Top
e o
From tthe Performance
r nlink. tra
v
Consumers -
EThe TopnoConsumers page has several tabs for displaying your database as a single-system
image. The Overview tabbed page contains four pie charts: Top Clients, Top Services, Top
Modules, and Top Actions. Each chart provides a different perspective regarding the top
resource consumers in your database.
The Top Services tabbed page displays performance-related information for the services that
are defined in your database. Using this page, you can enable or disable tracing at the service
level, as well as view the resulting SQL trace file.

Service Aggregation Configuration
Automatic service aggregation level of statistics

DBMS_MONITOR used for finer granularity of service

aggregations:
SERV_MOD_ACT_STAT_ENABLE
SERV_MOD_ACT_STAT_DISABLE
Possible additional aggregation levels:
SERVICE_NAME/MODULE s a
) ha
SERVICE_NAME/MODULE/ACTION m
o
c e
Tracing services, modules, and actions: y s
u n is uid
SERV_MOD_ACT_TRACE_ENABLE
b r n t G
SERV_MOD_ACT_TRACE_DISABLE
n t o@ tude
S
Database settings persist sacross c ime instance t h i s restarts.
a
n o us e
r t o n t
v e s e
n (e 2012,
r t o n le
e a b
n Ev nimportant
s fer statistics and wait events are automatically aggregated and
On each
e r to by service.
instance,
- tra You do not have to do anything to set this up, except connect with
v
collected
o n
Edifferentnconnect strings by using the services that you want to connect to. However, to
achieve a finer level of granularity of statistics collection for services, you must use the
SERV_MOD_ACT_STAT_ENABLE procedure in the DBMS_MONITOR package. This procedure
enables statistics gathering for additional hierarchical combinations of
SERVICE_NAME/MODULE and SERVICE_NAME/MODULE/ACTION. The
SERV_MOD_ACT_STAT_DISABLE procedure stops the statistics gathering that was turned on.
The enabling and disabling of statistics aggregation within the service applies to every
instance accessing the database. These settings are persistent across instance restarts.
The SERV_MOD_ACT_TRACE_ENABLE procedure enables tracing for services with three
hierarchical possibilities: SERVICE_NAME, SERVICE_NAME/MODULE, and
SERVICE_NAME/MODULE/ACTION. The default is to trace for all instances that access the
database. A parameter is provided that restricts tracing to specified instances where poor
performance is known to exist. This procedure also gives you the option of capturing relevant
waits and bind variable values in the generated trace files.
SERV_MOD_ACT_TRACE_DISABLE disables the tracing at all enabled instances for a given
combination of service, module, and action. Like the statistics gathering mentioned previously,
service tracing persists across instance restarts.

Service, Module, and Action Monitoring
For the ERP service, enable monitoring for the exceptions

pay action in the PAYROLL module.

EXEC DBMS_MONITOR.SERV_MOD_ACT_STAT_ENABLE(
service_name => 'ERP', module_name=> 'PAYROLL',
action_name => 'EXCEPTIONS PAY')
For the ERP service, enable monitoring for all the actions in
the PAYROLL module: a
EXEC DBMS_MONITOR.SERV_MOD_ACT_STAT_ENABLE(service_name) h
as
=>
'ERP', module_name=> 'PAYROLL', action_name => NULL); c om
i s ys ide
For the HOT_BATCH service, enable monitoring r u n for G uall
b en t
actions in the posting module: o @ tud
e nt S
EXEC DBMS_MONITOR.SERV_MOD_ACT_STAT_ENABLE(service_name
s c im this =>
'HOT_BATCH', module_name =>'POSTING',
na us e action_name => NULL);
n
rto se t o
n (e 2012,
r t o n le
e a b
n Ev performance
s f er data tracing for important modules and actions within each
You can
e r
service.
enable
t an
toThe -performance
r statistics are available in the V$SERV_MOD_ACT_STATS view.
v
EConsider o n
n the following actions, as implemented in the slide:
For the ERP service, enable monitoring for the exceptions pay action in the payroll
module.
Under the ERP service, enable monitoring for all the actions in the payroll module.
Under the HOT_BATCH service, enable monitoring for all actions in the posting module.
Verify the enabled service, module, action configuration with the SELECT statement below:
COLUMN AGGREGATION_TYPE FORMAT A21 TRUNCATED HEADING 'AGGREGATION'
COLUMN PRIMARY_ID FORMAT A20 TRUNCATED HEADING 'SERVICE'
COLUMN QUALIFIER_ID1 FORMAT A20 TRUNCATED HEADING 'MODULE'
COLUMN QUALIFIER_ID2 FORMAT A20 TRUNCATED HEADING 'ACTION'
SELECT * FROM DBA_ENABLED_AGGREGATIONS ;
The output might appear as follows:
AGGREGATION SERVICE MODULE ACTION
--------------------- -------------------- ---------- -------------
SERVICE_MODULE_ACTION ERP PAYROLL EXCEPTIONS
PAY
SERVICE_MODULE_ACTION ERP PAYROLL
SERVICE_MODULE_ACTION HOT_BATCH POSTING

Service Performance Views
Service, module, and action information in:

V$SESSION
V$ACTIVE_SESSION_HISTORY
Service performance in:
V$SERVICE_STATS
V$SERVICE_EVENT
V$SERVICE_WAIT_CLASS a
V$SERVICEMETRIC ha s
m )
V$SERVICEMETRIC_HISTORY o
c e
y s
V$SERV_MOD_ACT_STATS
u n is uid
DBA_ENABLED_AGGREGATIONS b r n t G
@ e
DBA_ENABLED_TRACES nto Stud
c i me this
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Emodule,s e action information are visible in V$SESSION and
fand
The service,
o
rt n-tra n
v e
V$ACTIVE_SESSION_HISTORY.
E no
The call times and performance statistics are visible in V$SERVICE_STATS,
V$SERVICE_EVENT, V$SERVICE_WAIT_CLASS, V$SERVICEMETRIC, and
V$SERVICEMETRIC_HISTORY.
When statistics collection for specific modules and actions is enabled, performance measures
are visible at each instance in V$SERV_MOD_ACT_STATS.
More than 600 performance-related statistics are tracked and visible in V$SYSSTAT. Of these,
28 statistics are tracked for services. To see the statistics measured for services, run the
following query: SELECT DISTINCT stat_name FROM v$service_stats
Of the 28 statistics, DB time and DB CPU are worth mentioning. DB time is a statistic that
measures the average response time per call. It represents the actual wall clock time for a call
to complete. DB CPU is an average of the actual CPU time spent per call. The difference
between response time and CPU time is the wait time for the service. After the wait time is
known, and if it consumes a large percentage of response time, then you can trace at the
action level to identify the waits.
Note: DBA_ENABLED_AGGREGATIONS displays information about enabled on-demand
statistic aggregation. DBA_ENABLED_TRACES displays information about enabled traces.

Quiz
Which of the following statements regarding Oracle Services is

not correct?
a. You can group work by type under services.
b. Users who share a service should have the same service-
level requirements.
c. You use DBMS_SERVICE to manage services, not srvctl
or Enterprise Manager. s a
) ha
c om
i s ys ide
r u n Gu
@ b ent
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o c
rt n-tra
v
E no c is not correct.
Statement

Quiz
Is the following statement regarding performance thresholds

true or false? The two performance thresholds that can be

explicitly set for each service are:
(a) SERVICE_ELAPSED_TIME: The response time for calls
(b) SERVICE_CPU_TIME: The CPU time for calls
a. True
b. False a
ha s
m )
o
c e
y s
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o a
rt n-tra
v
no is true.
E statement
The

Summary

Configure and manage services in a RAC environment

Use services with client applications
Use services with the Database Resource Manager
Use services with the Scheduler
Set performance-metric thresholds on services
s a
Configure services aggregation and tracing
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Practice 8 Overview

Creating and managing services using EM

Using server-generated alerts in combination with services
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

High Availability of Connections
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives

Configure client-side connect-time load balancing

Configure client-side connect-time failover
Configure server-side connect-time load balancing
Use the Load Balancing Advisory (LBA)
Describe the benefits of Fast Application Notification (FAN)
s a
Configure server-side callouts
) ha
Configure the server- and client-side ONS o m
y s c e
Configure Transparent Application Failover is id
un(TAF)Gu br ent
@
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfsee e
For more
r t o information,
t r a the following page:
Eve nhttp://www.oracle.com/technology/products/database/clustering/pdf/awmrac11g.pdf
on-

Types of Workload Distribution
Connection balancing is rendered possible by configuring

multiple listeners on multiple nodes:

Client-side connect-time load balancing
Client-side connect-time failover
Server-side connect-time load balancing
Runtime connection load balancing is rendered possible by
using connection pools: s a
Work requests automatically balanced across the pool of) h
a
connections c om
s y s de
Native feature of Oracle Universal Connection u n i u i
Pool (UCP)
for Java, and ODP.NET connection @ pool

br ent G
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nlisteners
s fe on multiple nodes can be configured to handle client connection
e o
With RAC, multiple
rt fornthe trasame database service.
v
requests -
no
EA multiple-listener configuration enables you to leverage the following failover and load-
balancing features:
Client-side connect-time load balancing
Client-side connect-time failover
Server-side connect-time load balancing
These features can be implemented either one by one, or in combination with each other.
Moreover, if you are using connection pools, you can benefit from readily available runtime
connection load balancing to distribute the client work requests across the pool of
connections established by the middle tier. This possibility is offered by the Oracle Universal
Connection Pool (UCP) for Java feature as well as Oracle Data Provider for .NET (ODP.NET)
connection pool.
Note: Starting with Oracle Database 11g Release 1 (11.1.0.7), Oracle has released the new
Universal Connection Pool for JDBC. Consequently, Oracle is deprecating the existing JDBC
connection pool (that is, Implicit Connection Cache) that was introduced in Oracle Database
10g Release 1.

Client-Side Connect-Time Load Balancing
ERP =
(DESCRIPTION =
(ADDRESS_LIST =
(LOAD_BALANCE=ON)
(ADDRESS=(PROTOCOL=TCP)(HOST=node1vip)(PORT=1521))
)
s a
)ha
Random m
co
s
access
u n isy uide
b r nt G
n t o@ tude
me is S
node1 sci node2th
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe load balancing feature enables clients to randomize connection
E connect-time
e o
The client-side
rt among traa list of available listeners. Oracle Net progresses through the list of protocol
v
requests n -
no in a random sequence, balancing the load on the various listeners. Without this
Eaddresses
feature, Oracle Net always takes the first protocol address to attempt a connection.
You enable this feature by setting the LOAD_BALANCE=ON clause in the corresponding client-
side TNS entry.
Note: For a small number of connections, the random sequence is not always even.

Client-Side Connect-Time Failover
ERP =
(DESCRIPTION =
(ADDRESS_LIST =
(LOAD_BALANCE=ON)
(FAILOVER=ON) 3
) 4
(CONNECT_DATA=(SERVICE_NAME=ERP))) s a
)ha
m
co
s
u n isy uide
b r nt G
1 n t o@ tude
c
node2vipi me this S
as use
2 nnode1vip

e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
This feature
Eenablesn s fe
clients to connect to another listener if the initial connection to the first
e o
rt fails. tranumber of listener protocol addresses in the connect descriptor determines
listener
v n -The
no listeners are tried. Without client-side connect-time failover, Oracle Net attempts a
Ehow many
connection with only one listener. As shown by the example in the slide, client-side connect-
time failover is enabled by setting FAILOVER=ON in the corresponding client-side TNS entry.
In the example, you expect the client to randomly attempt connections to either NODE1VIP or
NODE2VIP, because LOAD_BALANCE is set to ON. In the case where one of the nodes is
down, the client cannot know this. If a connection attempt is made to a down node, the client
needs to wait until it receives the notification that the node is not accessible, before an
alternate address in the ADDRESS_LIST is tried.
Therefore, it is highly recommended to use virtual host names in the ADDRESS_LIST of your
connect descriptors. If a failure of a node occurs (1), the virtual IP address assigned to that
node is failed over and brought online on another node in the cluster (2). Thus, all client
connection attempts are still able to get a response from the IP address, without the need to
wait for the operating system TCP/IP timeout (3). Therefore, clients get an immediate
acknowledgement from the IP address, and are notified that the service on that node is not
available. The next address in the ADDRESS_LIST can then be tried immediately with no
delay (4).
Note: If you use connect-time failover, do not set GLOBAL_DBNAME in your listener.ora
file.

Server-Side Connect-Time Load Balancing
ERP = (DESCRIPTION=
(ADDRESS_LIST=(LOAD_BALANCE=ON)(FAILOVER=ON)
(ADDRESS=(PROTOCOL=TCP)(HOST=node2vip)(PORT=1521)))
6 4 3 2
ERP started on both instances
Listener Listener
1
5 s a
1 1
)ha
PMON PMON m
co
s
ide
isy uNode2
*.REMOTE_LISTENER=RACDB_LISTENERS
Node1
u n
r nt G
b
RACDB_LISTENERS=
n t o@ tude
me this S
(DESCRIPTION=
c i
(ADDRESS=(PROTOCOL=tcp)(HOST=node1vip)(PORT=1521))
n as use
(ADDRESS=(PROTOCOL=tcp)(HOST=node2vip)(PORT=1521)))
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E you n s fe listeners distribute service connection requests across a RAC
e o
The slide shows
rt Here, how
raclient application connects to the ERP service. On the server side, the
tthe
v
cluster. n -
nois using the dynamic service registration feature. This allows the PMON process of
Edatabase
each instance in the cluster to register service performance information with each listener in
the cluster (1). Each listener is then aware of which instance has a particular service started,
as well as how that service is performing on each instance.
You configure this feature by setting the REMOTE_LISTENER initialization parameter of each
instance to a TNS name that describes the list of all available listeners. The slide shows the
shared entry in the SPFILE as well as its corresponding server-side TNS entry.
Depending on the load information, as computed by the Load Balancing Advisory, and sent
by each PMON process, a listener redirects the incoming connection request (2) to the
listener of the node where the corresponding service is performing the best (3).
In the example, the listener on NODE2 is tried first. Based on workload information
dynamically updated by PMON processes, the listener determines that the best instance is
the one residing on NODE1. The listener redirects the connection request to the listener on
NODE1 (4). That listener then starts a dedicated server process (5), and the connection is
made to that process (6).
Note: For more information, refer to the Net Services Administrators Guide.

Fast Application Notification: Overview
.NET app C app C app Java app Java app

ODP.NET OCI API ONS OCI ONS Java JDBC

API API
ONS ONS
AQ
s a
HA )ha
Proxy ONS m
app
Events
s co
n i sy DB u i de
Callout HA
CRS
HA
EMDr
b u t G
script Events Events
n Control
n t o@ tude
Callout
i m e is S
exec HA s
a c e th
n n
Events u s Node1
rto se t o
n (e 2012,
r t o n le
e a b
n Ev Notification
s f er
r t o
Fast Application
t r a n (FAN) enables end-to-end, lights-out recovery of applications and
ve balancing
Eload - based on real transaction performance in a RAC environment. With FAN, the
nonservice built in to Oracle Real Application Clusters 11g is extended to applications
continuous
and mid-tier servers. When the state of a database service changes, (for example, up, down,
or not restarting), the new status is posted to interested subscribers through FAN events.
Applications use these events to achieve very fast detection of failures, and rebalancing of
connection pools following failures, recovery, or planned changes. The easiest way to receive
all the benefits of FAN, with no effort, is to use a client that is integrated with FAN:
Oracle Universal Connection Pool (UCP) for Java
User extensible callouts
Connection Manager (CMAN)
Listeners
Oracle Notification Service (ONS) API
OCI Connection Pool or Session Pool
Transparent Application Failover (TAF)
ODP.NET Connection Pool
Note: Not all the preceding applications can receive all types of FAN events.

Fast Application Notification: Benefits
No need for connections to rely on connection timeouts

Used by Load Balancing Advisory to propagate load

information
Designed for enterprise application and management
console integration
Reliable distributed system that:
Detects high-availability event occurrences in a timely s a
manner ) ha
o m
Pushes notification directly to your applications ysc e
Tightly integrated with: u n is uid
b r n t G
Oracle JDBC applications using connection t o@ tudpools e
n
Enterprise Manager c i me this S
Data Guard Brokernna
s se
o t o u
r t
( e ve ense
o n n e2012,
Copyright
v e rt rab
n e
Eclientnorsfmid-tier
r t o
Traditionally,
t r a applications connected to the database have relied on
ve ontimeouts,
Econnection - out-of-band polling mechanisms, or other custom solutions to realize
n
that a system component has failed. This approach has huge implications in application
availability, because down times are extended and more noticeable.
With FAN, important high-availability events are pushed as soon as they are detected, which
results in a more efficient use of existing computing resources, and a better integration with
your enterprise applications, including mid-tier connection managers, or IT management
consoles, including trouble ticket loggers and email/paging servers.
FAN is, in fact, a distributed system that is enabled on each participating node. This makes it
very reliable and fault tolerant because the failure of one component is detected by another.
Therefore, event notification can be detected and pushed by any of the participating nodes.
FAN events are tightly integrated with Oracle Data Guard Broker, Oracle JDBC implicit
connection cache, ODP.NET, TAF, and Enterprise Manager. For example, Oracle Database
11g JDBC applications managing connection pools do not need custom code development.
They are automatically integrated with the ONS if implicit connection cache and fast
connection failover are enabled.
Note: For more information about FAN and Data Guard integration, refer to the lesson titled
Design for High Availability in this course.

FAN-Supported Event Types
Event type Description

SERVICE Primary application service
SRV_PRECONNECT Shadow application service event

(mid-tiers and TAF using primary and secondary
instances)
SERVICEMEMBER Application service on a specific instance
DATABASE Oracle database

s a
INSTANCE Oracle instance ) ha
o m
Oracle ASM instance y s c e
ASM
u n is uid
Oracle cluster node b r n t G
o@ tude
NODE
n
e is S t
SERVICEMETRICS Load Balancing
c i mAdvisory th
a s e
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n
v
Eevents s f er
r t o
FAN delivers
t r a n pertaining to the list of managed cluster resources shown in the slide.
vetableodescribes
EThe n- each of the resources.
n
Note: SRV_PRECONNECT and SERVICEMETRICS are discussed later in this lesson.

FAN Event Status
Event status Description

up Managed resource comes up.
down Managed resource goes down.
preconn_up Shadow application service comes up.
Shadow application service goes down.

preconn_down
s a
nodedown Managed node goes down. ) ha
o m
not_restarting Managed resource cannot fail over to ayremote s c e
node. u n is uid
b G
r locallynafter
t
restart_failed Managed resource fails to
t o @ tude a
start
i m en is S
discrete number of retries.
Unknown
a s c e th
Status is unrecognized.
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n Ev nsthe ferevent status for each of the managed cluster resources seen
e r to -tra
This table describes
Ev non
previously.

FAN Event Reasons
Event Reason Description

user User-initiated commands, such as srvctl and sqlplus
failure Managed resource polling checks detecting a failure
dependency Dependency of another managed resource that triggered

a failure condition
unknown Unknown or internal application state when event is
triggered s a
autostart Initial cluster boot: Managed resource has profile attribute )ha
AUTO_START=1, and was offline before the last Oracle m
co
s
Clusterware shutdown.
u n isy uide
Boot Initial cluster boot: Managed resource
b n G before
r was running
t
the last Oracle Clusterware shutdown.
n t o@ tude
PUBLIC_NW_DOWN The node is up, but a downed
m e inetwork s S prevents
c i h
connectivity. s
n a use t
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E for n s fe managed resource is associated with an event reason. The reason
e o
The event status
rt describes each
trawhat triggered the event. The table in the slide gives you the list of possible
v
further n -
o a corresponding description.
Ereasonsnwith

FAN Event Format
<Event_Type>
VERSION=<n.n>
[service=<serviceName.dbDomainName>]
[database=<dbName>] [instance=<sid>]
[host=<hostname>]
status=<Event_Status>
reason=<Event_Reason>
[card=<n>]
timestamp=<eventDate> <eventTime> s a
) ha
o m
SERVICE VERSION=1.0 service=ERP.oracle.com
y s c e
database=RACDB status=up reason=user card=4
u n is uid
timestamp=16-Mar-2004 19:08:15 b r n t G
n t o@ tude
NODE VERSION=1.0 host=strac-1
c i me this S
status=nodedown timestamp=16-Mar-2004
n as use 17:35:53
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eits type, s e
fstatus,
r t o
In addition to
t r a n and reason, a FAN event has other payload fields to further
ve othen-unique cluster resource whose status is being monitored and published:
Edescribe
The n event payload version, which is currently 1.0
The name of the primary or shadow application service. This name is excluded from
NODE events.
The name of the RAC database, which is also excluded from NODE events
The name of the RAC instance, which is excluded from SERVICE, DATABASE, and
NODE events
The name of the cluster host machine, which is excluded from SERVICE and
DATABASE events
The service cardinality, which is excluded from all events except for SERVICE status=up
events
The server-side date and time when the event is detected
The general FAN event format is described in the slide along with possible FAN event
examples. Note the differences in event payload for each FAN event type.

Load Balancing Advisory: FAN Event
Parameter Description
Version Version of the event payload
Event type SERVICEMETRICS
Service Matches DBA_SERVICES
Database unique name Unique DB name supporting the service

Time stamp Date and time stamp (local time zone)
s a
)h a
Repeated m
Instance Instance name supporting the service ys
co e
u n is uid
Percent
Percentage of work to sendbtor this database
n t G and instance
Flag GOOD, VIOLATING,nNO t o@ u de
DATA,tUNKNOWN
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsAdvisoryfe
The Load
r t o Balancing
t r a FAN event is described in the slide. Basically, it contains a
ve onpercentage
Ecalculated - of work requests that should be sent to each instance. The flag
indicatesn the behavior of the service on the corresponding instance relating to the thresholds
set on that instance for the service.
Here is an example:
Notification Type: database/event/servicemetrics/prod
VERSION=1.0 database=PROD service=myServ { {instance=PROD2
percent=38 flag=GOOD aff=TRUE}{instance=PROD3 percent=62 flag=GOOD
aff=TRUE} } timestamp=2010-07-30 08:47:06
Note: Applications using Oracle Database 11g and UCP can take advantage of the affinity
feature. If the affinity flag is turned on in the Load Balancing Advisory event, then UCP
creates an Affinity Context for the web session such that when that session does a get
connection from the pool, the pool always tries to give it a connection to the instance it
connected to the first time it acquired a session. The choice of instance for the first connection
is based on the current load balancing advisory information. The affinity hint is automatic
when load balancing advisory is turned on through setting the goal on the service. This flag is
for temporary affinity that lasts for the duration of a web session.

Server-Side Callouts Implementation
The callout directory:

<Grid Home>/racg/usrco
Can store more than one callout
Grants execution on callout directory and callouts only to the
Oracle Clusterware user
Callout execution order is nondeterministic.
Writing callouts involves: s a
)ha
1. Parsing callout arguments: The event payload m
2. Filtering incoming FAN events s co
3. Executing event-handling programs u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E event s e
fdetected
r o
Each tdatabase
t r a n by the RAC High Availability (HA) framework results in the
ve oofn-each executable script or program deployed in the standard Oracle Clusterware
Eexecution
callout n
directory. On UNIX, it is <Grid Home>/racg/usrco. Unless your Oracle
Clusterware home directory is shared across the network, you must deploy each new callout
on each RAC node. The order in which these callouts are executed is nondeterministic.
However, RAC guarantees that all callouts are invoked once for each recognized event in an
asynchronous fashion. Thus, it is recommended to merge callouts whose executions need to
be in a particular order.
You can install as many callout scripts or programs as your business requires, provided each
callout does not incur expensive operations that delay the propagation of HA events. If many
callouts are going to be written to perform different operations based on the event received, it
might be more efficient to write a single callout program that merges each single callout.
Writing server-side callouts involves the steps shown in the slide. In order for your callout to
identify an event, it must parse the event payload sent by the RAC HA framework to your
callout. After the sent event is identified, your callout can filter it to avoid execution on each
event notification. Then, your callout needs to implement a corresponding event handler that
depends on the event itself and the recovery process required by your business.
Note: As a security measure, make sure that the callout directory and its contained callouts
have write permissions only to the system user who installed Oracle Clusterware.

Server-Side Callout Parse: Example
#!/bin/sh
NOTIFY_EVENTTYPE=$1
for ARGS in $*; do

PROPERTY=`echo $ARGS | $AWK -F"=" '{print $1}'`
VALUE=`echo $ARGS | $AWK -F"=" '{print $2}'`
case $PROPERTY in
VERSION|version) NOTIFY_VERSION=$VALUE ;;
SERVICE|service) NOTIFY_SERVICE=$VALUE ;;
DATABASE|database) NOTIFY_DATABASE=$VALUE ;;
INSTANCE|instance) NOTIFY_INSTANCE=$VALUE ;;
s a
HOST|host) NOTIFY_HOST=$VALUE ;;
) ha
STATUS|status) NOTIFY_STATUS=$VALUE ;; om
REASON|reason) NOTIFY_REASON=$VALUEys;; c e
u n is uid;;
CARD|card)
b r
NOTIFY_CARDINALITY=$VALUE
n t G;;
o@ tude
TIMESTAMP|timestamp) NOTIFY_LOGDATE=$VALUE
??:??:??) n t S
NOTIFY_LOGTIME=$PROPERTY
e ;;
esac
s c im this
done na se n u
r t o t o
( e ve ense
o n n e2012,
Copyright
v e rt rab
n Ewant your s e
fcallouts
Unless
r t oyou
t r a n to be executed on each event notification, you must first
ve the
Eidentify - parameters that are passed automatically to your callout during its execution.
event
non
The example in the slide shows you how to parse these arguments by using a sample Bourne
shell script.
The first argument that is passed to your callout is the type of event that is detected. Then,
depending on the event type, a set of PROPERTY=VALUE strings are passed to identify exactly
the event itself.
The script given in the slide identifies the event type and each pair of PROPERTY=VALUE
strings. The data is then dispatched into a set of variables that can be used later in the callout
for filtering purposes.
As mentioned in the previous slide, it might be better to have a single callout that parses the
event payload, and then executes a function or another program on the basis of information in
the event, as opposed to having to filter information in each callout. This becomes necessary
only if many callouts are required.
Note: Make sure that executable permissions are set correctly on the callout script.

Server-Side Callout Filter: Example
if ((( [ $NOTIFY_EVENTTYPE = "SERVICE" ] ||

[ $NOTIFY_EVENTTYPE = "DATABASE" ] || \
[ $NOTIFY_EVENTTYPE = "NODE" ] \
) && \
( [ $NOTIFY_STATUS = "not_restarting" ] || \
[ $NOTIFY_STATUS = "restart_failed" ] \
)) && \
( [ $NOTIFY_DATABASE = "HQPROD" ] || \
[ $NOTIFY_SERVICE = "ERP" ] \
s a
))
) ha
then o m
/usr/local/bin/logTicket $NOTIFY_LOGDATEys\ c e
$NOTIFY_LOGTIME u n is \ uid
b r n t G
$NOTIFY_SERVICE
o @ d e \
e nt
$NOTIFY_DBNAME
S tu \
s c e t his
im$NOTIFY_HOST
fi na sn u
r t o t o
( e ve ense
o n n e2012,
Copyright
v e rt rab
n fe shows you a way to filter FAN events from a callout script. This
Ein thensslide
e o
The example
tra on the example in the previous slide.
rt isnbased
v
example -
E no
Now that the event characteristics are identified, this script triggers the execution of the
trouble-logging program /usr/local/bin/logTicket only when the RAC HA framework
posts a SERVICE, DATABASE, or NODE event type, with a status set to either
not_restarting or restart_failed, and only for the production HQPROD RAC database
or the ERP service.
It is assumed that the logTicket program is already created and that it takes the arguments
shown in the slide.
It is also assumed that a ticket is logged only for not_restarting or restart_failed
events, because they are the ones that exceeded internally monitored timeouts and seriously
need human intervention for full resolution.

Server-Side ONS
localport=6100 # line added by Agent

remoteport=6200 # line added by Agent
nodes=node1:6200, node2:6200 # line added by Agent
s a
)h a
ONS ONS om
y s c e
OCR u n is uid
b
r ons.config
n t G
ons.config
n t o@ tude
Node1
c i me thNode2 is S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
The ONS n E nsfeis controlled by the <Grid HOME>/opmn/conf/ons.config
configuration
e o traThis file is automatically created during installation. Starting with Oracle
rt n-file.
v
configuration
no11g Release 2 (11.2) it is automatically maintained by the CRS ONS agent using
EDatabase
information stored in the OCR. There are three important parameters that are always
configured for each ONS:
The first is localport, the port that ONS uses to talk to local clients.
The second is remoteport, the port that ONS uses to talk to other ONS daemons.
The third parameter is called nodes. It specifies the list of other ONS daemons to talk
to. This list includes all RAC ONS daemons, and all mid-tier ONS daemons. Node
values are given as either host names or IP addresses followed by their remoteport.
This information is stored in Oracle Cluster Registry (OCR).
In the slide, it is assumed that ONS daemons are already started on each cluster node. This
should be the default situation after a correct RAC installation.

Optionally Configuring the Client-Side ONS
Mid-tier1
ons.config ONS $ onsctl start 2
localport=6100
remoteport=6200 1
nodes=node1:6200,node2:6200
s a
)ha
co m
s
ONS
u n
y
isONS u ide
b r nt G
OCR
t o @ tude
n
ons.config
c i me this S ons.config
Node1
n as use Node2
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E 11g s e
fRelease
Oracle
r t o
Database
t r a n 2 introduces a new set of APIs for Oracle RAC Fast
ve onNotification
EApplication - (FAN) events. These APIs provide an alternative for taking advantage
n
of the high-availability (HA) features of Oracle Database, if you do not use Universal
Connection Pool or Oracle JDBC connection caching. These APIs are not a part of Oracle
JDBC APIs. For using Oracle RAC Fast Application Notification, the simplefan.jar file
must be present in the CLASSPATH, and either the ons.jar file must be present in the
CLASSPATH or an Oracle Notification Services (ONS) client must be installed and running in
the client system. To use ONS on the client side, you must configure all the RAC nodes in the
ONS configuration file. A sample configuration file might look like the one shown in the slide.
After configuring ONS, you start the ONS daemon with the onsctl start command. It is
your responsibility to make sure that an ONS daemon is running at all times. You can check
that the ONS daemon is active by executing the onsctl ping command.
Note: With Oracle Database 10g Release 2 and later, there is no requirement to use ONS
daemons on the mid-tier when using the Oracle Universal Connection Pool. To configure this
option, use either the OracleDataSource property or a setter API setONSConfiguration
(configStr). The input to this API is the contents of the ons.config file specified as a
string. The ons.jar file must be on the clients CLASSPATH. There are no daemons to start
or manage.

UCP JDBC Fast Connection Failover: Overview
Service Mid-tier1
Service or node
UP event ONS
DOWN event
UCP JDBC
Event
handler
Connections Connections
reconnected Connection Cache marked down &
s a
cleaned up
) ha
Connections Connections om
using Listeners usingysc e
service names service i s
n names u i d
Connections r
b en u t G
load balancing
t o @ ud
ONS en ONSSt
s c im this
Node1 n a use Noden
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Oracleo n
Universal n fe Pool (UCP) provides a tight integration with Oracle RAC
E Connection
s
e
database
v n - tra like Fast Connection Failover (FCF). Basically, FCF is a FAN client
rt features
no through the connection pool. FCF quickly and automatically recovers lost or
Eimplemented
damaged connections. This automatic connection management results from FAN events
received by the local ONS daemon, or by a remote ONS if a local one is not used, and
handled by a special event handler thread. Both JDBC thin and JDBC OCI drivers are
supported.
Therefore, if UCP and FCF are enabled, your Java program automatically becomes an ONS
subscriber without having to manage FAN events directly.
Whenever a service or node down event is received by the mid-tier ONS, the event handler
automatically marks the corresponding connections as down and cleans them up. This
prevents applications that request connections from the cache from receiving invalid or bad
connections.
Whenever a service up event is received by the mid-tier ONS, the event handler recycles
some unused connections, and reconnects them using the event service name. The number
of recycled connections is automatically determined by the connection cache. Because the
listeners perform connection load balancing, this automatically rebalances the connections
across the preferred instances of the service without waiting for connection requests or
retries.
For more information see Oracle Universal Connection Pool for JDBC Developers Guide.
Note: Similarly, ODP.NET also allows you to use FCF using AQ for FAN notifications.

Using Oracle Streams Advanced Queuing for FAN
Use AQ to publish FAN to ODP.NET and OCI.

Turn on FAN notification to an alert queue.

exec DBMS_SERVICE.MODIFY_SERVICE (
service_name => 'SELF-SERVICE', aq_ha_notification => TRUE);
View published FAN events:

SQL> select object_name,reason
2 from dba_outstanding_alerts; s a
)ha
OBJECT_NAME REASON m
co
s
isy uide
----------- ----------------------------------------
xwkE Database xwkE (domain ) up as of time
u n
r nt G
2005-12-30 11:57:29.000000000 -05:00; b
reason code: user
n t o@ tude
JFSERV Composite service xwkE up
me asthiof Stime
s-05:00;
c i
n as use
2006-01-02 05:27:46.000000000
ton e to
reason code: user
e r
v ens
( e
o n n e2012,
Copyright
v e rt rab
n E Clusters
n s fe publish FAN events to a system alert queue in the database using
e o
Real Application
rt StreamstraAdvanced Queuing (AQ). ODP.NET and OCI client integration uses this
v
Oracle n -
Emethodntoosubscribe to FAN events.
To have FAN events for a service posted to that alert queue, the notification must be turned
on for the service using either the DBMS_SERVICE PL/SQL package as shown in the slide, or
by using the Enterprise Manager interface.
To view FAN events that are published, you can use the DBA_OUTSTANDING_ALERTS or
DBA_ALERT_HISTORY views. An example using DBA_OUTSTANDING_ALERTS is shown in
the slide.

JDBC/ODP.NET FCF Benefits
Database connections are balanced across preferred

instances according to LBA.

Database work requests are balanced across preferred
instances according to LBA.
Database connections are anticipated.
Database connection failures are immediately detected
and stopped. s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EFCF, nyour
s feexisting Java applications connecting through Oracle JDBC and
e o
By enabling
tra or your .NET applications using ODP.NET connection pools and
rt nservices,
v
application -
no services benefit from the following:
Eapplication
All database connections are balanced across all RAC instances that support the new
service name, instead of having the first batch of sessions routed to the first RAC
instance. This is done according to the Load Balancing Advisory algorithm you use (see
the next slide). Connection pools are rebalanced upon service, instance, or node up
events.
The connection cache immediately starts placing connections to a particular RAC
instance when a new service is started on that instance.
The connection cache immediately shuts down stale connections to RAC instances
where the service is stopped on that instance, or whose node goes down.
Your application automatically becomes a FAN subscriber without having to manage
FAN events directly by just setting up flags in your connection descriptors.
An exception is immediately thrown as soon as the service status becomes
not_restarting, which avoids wasteful service connection retries.
Note: For more information about how to subscribe to FAN events, refer to the Oracle
Database JDBC Developers Guide and Oracle Data Provider for .NET Developers Guide.

Load Balancing Advisory
The Load Balancing Advisory (LBA) is an advisory for

sending work across RAC instances.

The LBA advice is available to all applications that send
work:
JDBC and ODP connection pools
Connection load balancing
The LBA advice sends work to where services are s a
executing well and resources are available: )ha
co m
Relies on service goodness s
Adjusts distribution for different power nodes, u n
y
isdifferentu ide
b r nt G
priority and shape workloads, and changing
t o @ tuddemand e
Stops sending work to slow,m n
e orisfailed
hung, S nodes
c i h
n as use t
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe use persistent connections that span the instances of RAC offering a
Eapplications
e o
Well-written
tra are created infrequently and exist for a long duration. Work comes into
rt Connections
v
service. n -
no with high frequency, borrows these connections, and exists for a relatively short
Ethe system
duration.
The Load Balancing Advisory has the task of advising the direction of incoming work to the
RAC instances that provide optimal quality of service for that work. The LBA algorithm uses
metrics sensitive to the current performance of services across the system.
The Load Balancing Advisory is deployed with Oracles key clients, such as Connection Load
Balancing, Oracle Universal Connection Pool, OCI Session Pool, Oracle Data Provider (ODP)
Connection Pool for .NET, and is open for third-party subscription via ONS.
Using the Load Balancing Advisory for load balancing recognizes machine power differences,
sessions that are blocked in wait, failures that block processing, as well as competing
services of different importance. Using the Load Balancing Advisory prevents sending work to
nodes that are overworked, hung, or failed.

UCP JDBC/ODP.NET Runtime Connection
Load Balancing: Overview
CRM work requests
Connection Cache
10%
?
s a
60%
30%
)h a
m
co
s
u n isy uide
b r nt G
RAC RAC en
to@ Stude RAC
i m
c e th i s
Inst1 CRM is CRM is
n a s
Inst2
s CRM is Inst3
very busy.
o n
not busy.
o u busy.
r t t
( e ve ense
o n n e2012,
Copyright
v e rt rab
n E the nLoad
s feBalancing Advisory, work requests to RAC instances are assigned on
Without
e o using
rt basis, trawhich is suitable when each instance is performing equally well. However, if
avrandom n -
Eone of ntheoinstances becomes more burdened than the others because of the amount of work
resulting from each connection assignment, the random model does not perform optimally.
The Runtime Connection Load Balancing feature provides assignment of connections based
on the Load Balancing Advisory information from the instances in the RAC cluster. The
Connection Cache assigns connections to clients on the basis of a relative number indicating
what percentage of work requests each instance should handle.
In the diagram in the slide, the feedback indicates that the CRM service on Inst1 is so busy
that it should service only 10% of the CRM work requests; Inst2 is so lightly loaded that it
should service 60%; and Inst3 is somewhere in the middle, servicing 30% of requests. Note
that these percentages apply to, and the decision is made on, a per-service basis. In this
example, CRM is the service in question.
Note: Runtime Connection Load Balancing is a feature of Oracle connection pools.

Connection Load Balancing in RAC
This is also known as server-side load balancing.

Connection load balancing allows listeners to distribute

connection requests to the best instances.
Three metrics are available for listeners to decide:
Session count by instance
Run queue length of the node
Service goodness s a
ha
The metric used depends on the connection load- om)
balancing goal defined for the service: y s c e
LONG u n is uid
b r n t G
NONE
n t o@ tude
SHORT c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n e
E nsfrequests
Balancing
r t o connection
t r a is referred to as connection load balancing. Connections are
ve toodifferent
Erouted n- instances in the cluster on the basis of load information available to the
listener.nThree metrics are available for the listeners to use when selecting the best instance:
Session count by instance: For services that span RAC instances uniformly and
similar capacity nodes, the session count evenly distributes the sessions across RAC.
This method is used when the services connection load-balancing goal is set to LONG.
Run queue length of the node: For services that use a subset of RAC instances and
different capacity nodes, the run queue length places more sessions on the node with
least load at the time of connection creation.
Goodness by service: The goodness of the service is a ranking of the quality of service
that the service is experiencing at an instance. It also considers states such as restricted
access to an instance. This method is used when the services connection load-
balancing goal is set to SHORT. To prevent a listener from routing all connections to the
same instance between updates to the goodness values, each listener adjusts its local
ratings by a delta as connections are distributed. The delta value used represents an
average of resource time that connections consume by using a service. To further
reduce login storms, the listener uses a threshold delta when the computed delta is too
low because no work was sent over the connections yet.

Load Balancing Advisory: Summary
Uses DBMS_SERVICE.GOAL
Service time: weighted moving average of elapsed time

Throughput: weighted moving average of throughput
AWR
Calculates goodness locally (MMNL), forwards to master MMON
Master MMON builds advisory for distribution of work across RAC,
and posts load balancing advice to AQ
IMON retrieves advice and send it to ONS s a
EMON retrieves advice and send it to OCI ) ha
o m
Local MMNL post goodness to PMON sc
uni
Listeners use DBMS_SERVICE.CLB_GOAL=SHORTs y
u ide
Use goodness from PMON to distribute @ br ent G
connections.
Load Balancing Advisory users (inside e ntothe pools) S tud
Use percentages and flags stoc im work.
send e t his
n na us
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Ethe Load s e
fBalancing
r t o
You enable
t r a n Advisory when setting the services goal to
ve on-
EDBMS_SERVICE.GOAL_SERVICE_TIME or to DBMS_SERVIVCE.GOAL_THROUGHPUT.
n
MMNL (Manageability MoNitor Light) calculates the service metrics for service goal and
resource consumption every five seconds. MMNL derives the service goodness from this
data. MMON computes and posts the LBA FAN event to a system queue, and MMNL
forwards the service goodness and delta to PMON.
IMON (Instance Monitor) and EMON (Event MONitor) retrieve the event from the queue, and
PMON forwards the goodness and delta values to the listeners.
IMON posts the LBA FAN event to the local ONS daemon, and EMON posts it to AQ
subscribers.
The server ONS sends the event to the mid-tier ONS (if used).
The mid-tier receives the event and forwards them to subscribers. Each connection pool
subscribes to receive events for its own services. On receipt of each event, the Connection
Pool Manager refreshes the ratio of work to forward to each RAC instance connection part of
the pool. It also ranks the instances to use when aging out connections.
Work requests are routed to RAC instances according to the ratios calculated previously.

Monitoring LBA FAN Events
SQL> SELECT TO_CHAR(enq_time, 'HH:MI:SS') Enq_time, user_data

2 FROM sys.sys$service_metrics_tab
3 ORDER BY 1 ;
ENQ_TIME USER_DATA
-------- -----------------------------------------------------

04:19:46 SYS$RLBTYP('JFSERV', 'VERSION=1.0 database=xwkE
s a
service=JFSERV { {instance=xwkE2 percent=50
flag=UNKNOWN}{instance=xwkE1 percent=50 flag=UNKNOWN} )ha
m
co
} timestamp=2006-01-02 06:19:46')
s
04:20:16 SYS$RLBTYP('JFSERV', 'VERSION=1.0 database=xwkE
u n isy uide
service=JFSERV { {instance=xwkE2 percent=80
b r nt G
o@ tude
flag=UNKNOWN}{instance=xwkE1 percent=20 flag=UNKNOWN}
} timestamp=2006-01-02 06:20:16') n t
SQL>
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Ethe SQL s e
fquery
You can
r t o use
t r a n shown in the slide to monitor the Load Balancing Advisory FAN
ve foroeach
Eevents - of your services.
n n

FAN Release Map
Oracle FAN Product Event FAN Event Received and Used

Release Integration System
10.1.0.2 UCP&FCF ONS Up/down/Load Balancing Advisory

Server-side callouts RAC Up/down
10.1.0.3 CMAN ONS Down

Listeners PMON Up/down/LBA
10.2 OCI connection pool AQ Down

s a
OCI session pool AQ Down
) ha
TAF AQ Down o m
y s c e
ODP.NET AQ Up/down/LBA
u n is uid
Custom OCI AQ All
b r n t G
DG Broker (10.2.0.3) AQ Down
n t o@ tude
11.1 OCI session pool AQ
c i meLBAthis S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Emap for
n s fe is shown in the slide.
e o
The release FAN
tra for Load Balancing Advisory.
rtLBAnstands
v
Note: -
E no

Transparent Application Failover: Overview
TAF Basic TAF Preconnect

2 Application 2 Application
OCI Library 6 OCI Library 6
4 Net Services 4 Net Services
3 7 3
5 5
s a
)ha
8 m
co
7
AP 3 ERP 3 s
u n is3 y uide
1
7
1
b r nt G
n t o@ tude
AP
c i sS
meERPthiERP_PRECONNECT
n as use 1
t o n t o
e v er nse
n ( 2012,
r n
to able
e
Ev nsferof the OCI driver. It enables your application to automatically
TAF istoanruntimeafeature
v e r ton-the
reconnect tr service if the initial connection fails. During the reconnection, although your
no
Eactive transactions are rolled back, TAF can optionally resume the execution of a SELECT
statement that was in progress. TAF supports two failover methods:
With the BASIC method, the reconnection is established at failover time. After the
service has been started on the nodes (1), the initial connection (2) is made. The listener
establishes the connection (3), and your application accesses the database (4) until the
connection fails (5) for any reason. Your application then receives an error the next time
it tries to access the database (6). Then, the OCI driver reconnects to the same service
(7), and the next time your application tries to access the database, it transparently uses
the newly created connection (8). TAF can be enabled to receive FAN events for faster
down events detection and failover.
The PRECONNECT method is similar to the BASIC method except that it is during the
initial connection that a shadow connection is also created to anticipate the failover. TAF
guarantees that the shadow connection is always created on the available instances of
your service by using an automatically created and maintained shadow service.
Note: Optionally, you can register TAF callbacks with the OCI layer. These callback functions
are automatically invoked at failover detection and allow you to have some control of the
failover process. For more information, refer to the Oracle Call Interface Programmers Guide.

TAF Basic Configuration Without FAN: Example
$ srvctl add service -d RACDB -s apsvc -r orcl1,orcl3 \

-P BASIC -e select -m basic -z 10 -w 2

$ srvctl start service -d orcl -s apsvc
apsvc =
(DESCRIPTION =
(ADDRESS = (PROTOCOL = TCP)(HOST = cluster01-scan)
(PORT = 1521))
(CONNECT_DATA =
(SERVICE_NAME = apsvc)))
s a
)ha
m
co
$ sqlplus AP/AP@apsvc
s
SQL> select inst_id,username,service_name,failover_type,
u n isy uide
failover_method from gv$session where username='AP';
b r nt G
n t o@ tuFAILOVER_M de
me this S ----------
INST_ID USERNAME SERVICE_NAME FAILOVER_TYPE
------- -------- ------------ c------------- i
1 AP n
apsvcas use SELECT BASIC
n
rto se t o
n (e 2012,
r t o n le
e a b
n
v
ETAF, s f er
Before
r t o
using
t r a n recommended that you create and start a service that to be used when
it is
ve onconnections.
Eestablishing - By doing so, you benefit from the integration of TAF and services.
When youn wish to use BASIC TAF with a service, you should use the -P BASIC option when
creating the service. After the service is created, you simply start it on your database. TAF
can be configured at the client side in tnsnames.ora or at the server side using the srvctl
utility as shown above. Configuring it at the server is preferred as it is convenient to put the
configuration in a single place (the server).
Your application needs to connect to the service by using a connection descriptor similar to
the one shown in the slide. In the example above, notice that the cluster SCAN is used in the
descriptor. Once connected, the GV$SESSION view will reflect that the connection is TAF-
enabled. The FAILOVER_METHOD and FAILOVER_TYPE column reflects this and confirms
the TAF configuration is correct.

TAF Basic Configuration with FAN: Example
$ srvctl add service -d RACDB -s AP -r I1,I2

$ srvctl start service -d RACDB -s AP
execute dbms_service.modify_service ( ,-
service_name => 'AP' ,-
aq_ha_notifications => true ,-
failover_method => dbms_service.failover_method_basic ,-
s a
failover_type => dbms_service.failover_type_session
failover_retries => 180, failover_delay => 5
,-
,- )ha
clb_goal => dbms_service.clb_goal_long); m
co
s
u n isy uide
AP =
b r nt G
o@ tude
(DESCRIPTION =(FAILOVER=ON)(LOAD_BALANCE=ON)
(ADDRESS=(PROTOCOL=TCP)(HOST=N1VIP)(PORT=1521)) n t
(ADDRESS=(PROTOCOL=TCP)(HOST=N2VIP)(PORT=1521))
c i me this S
(CONNECT_DATA = (SERVICE_NAME
n as =uAP))) s e
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E 10g s e
fRelease
Oracle
r t o
Database
t r a n 2 supports server-side TAF with FAN. To use server-side TAF,
ve and
Ecreate - your service using SRVCTL, then configure TAF in the RDBMS by using the
start
non package as shown in the slide. When done, make sure that you define a
DBMS_SERVICE
TNS entry for it in your tnsnames.ora file. Note that this TNS name does not need to
specify TAF parameters as with the previous slide.

TAF Preconnect Configuration: Example
$ srvctl add service -d RACDB -s ERP -r I1 a I2 \

> -P PRECONNECT
$ srvctl start service -d RACDB -s ERP
ERP =
a
has
(CONNECT_DATA = (SERVICE_NAME = ERP)
(FAILOVER_MODE = (BACKUP=ERP_PRECONNECT)
m )
c o
(TYPE=SESSION)(METHOD=PRECONNECT))))
de n i sys ui
ERP_PRECONNECT =
b ru nt G
t o @ tude
(ADDRESS=(PROTOCOL=TCP)(HOST=N1VIP)(PORT=1521)) n
e is S
c i m th
a s e
(CONNECT_DATA = (SERVICE_NAME
o n n o us = ERP_PRECONNECT)))
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n EvPRECONNECT
s f er TAF, it is recommended that you create a service with preferred
In order
e r to use
an Also, in order for the shadow service to be created and managed
to -tinstances.
r
v
and available
non by Oracle Clusterware, you must define the service with the P PRECONNECT
Eautomatically
option. The shadow service is always named using the format
<service_name>_PRECONNECT.
As with the BASIC method without FAN, you need to use a special connection descriptor to
use the PRECONNECT method while connecting to the service. One such connection
descriptor is shown in the slide.
The main differences with the previous example are that METHOD is set to PRECONNECT and
an addition parameter is added. This parameter is called BACKUP and must be set to another
entry in your tnsnames.ora file that points to the shadow service.
Note: In all cases where TAF cannot use the PRECONNECT method, TAF falls back to the
BASIC method automatically.

TAF Verification
SELECT machine, failover_method, failover_type,
failed_over, service_name, COUNT(*)
FROM v$session
GROUP BY machine, failover_method, failover_type,
failed_over, service_name;
MACHINE FAILOVER_M FAILOVER_T FAI SERVICE_N COUNT(*)

1st ------- ---------- ---------- --- -------- --------
node node1 BASIC SESSION NO AP 1
node1 PRECONNECT SESSION NO ERP 1
s a
MACHINE FAILOVER_M FAILOVER_T FAI SERVICE_N COUNT(*) )ha
2nd ------- ---------- ---------- --- --------- -------- m
co
node node2 s
NONE NONE NO ERP_PRECO
u n isy 1 uide
G
br ent COUNT(*)
MACHINE FAILOVER_M FAILOVER_T FAI@ SERVICE_N
2nd
------- ---------- ----------en to --------
--- S t ud --------
node
after
node2 BASIC SESSION
s c im YES
e t hisAP 1
node2 PRECONNECT SESSION a
n o us YES ERP_PRECO 1
r t o n t
v e s e
n (e 2012,
r t o n le
e a b
n Evwhether s f er
r t o
To determine
t r a n TAF is correctly configured and that connections are associated with a
ve option,
Efailover - you can examine the V$SESSION view. To obtain information about the
connectednonclients and their TAF status, examine the FAILOVER_TYPE, FAILOVER_METHOD,
FAILED_OVER, and SERVICE_NAME columns. The example includes one query that you
could execute to verify that you have correctly configured TAF. This example is based on the
previously configured AP and ERP services, and their corresponding connection descriptors.
The first output in the slide is the result of the execution of the query on the first node after two
SQL*Plus sessions from the first node have connected to the AP and ERP services,
respectively. The output shows that the AP connection ended up on the first instance.
Because of the load-balancing algorithm, it can end up on the second instance. Alternatively,
the ERP connection must end up on the first instance because it is the only preferred one.
The second output is the result of the execution of the query on the second node before any
connection failure. Note that there is currently one unused connection established under the
ERP_PRECONNECT service that is automatically started on the ERP available instance.
The third output is the one corresponding to the execution of the query on the second node
after the failure of the first instance. A second connection has been created automatically for
the AP service connection, and the original ERP connection now uses the preconnected
connection.

FAN Connection Pools and TAF Considerations
Both techniques are integrated with services and provide

service connection load balancing.

Do not use FCF when working with TAF, and vice versa.
Connection pools that use FAN are always preconnected.
TAF may rely on operating system (OS) timeouts to detect
failures.
FAN never relies on OS timeouts to detect failures. a
as )h
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe load balancing is a listener functionality, both FCF and TAF
Econnection
Because
e o the
ra from connection load balancing for services.
rt n-tbenefit
v
automatically
no use FCF, there is no need to use TAF. Moreover, FCF and TAF cannot work
EWhen you
together.
For example, you do not need to preconnect if you use FAN in conjunction with connection
pools. The connection pool is always preconnected.
With both techniques, you automatically benefit from VIPs at connection time. This means
that your application does not rely on lengthy operating system connection timeouts at
connect time, or when issuing a SQL statement. However, when in the SQL stack, and the
application is blocked on a read/write call, the application needs to be integrated with FAN in
order to receive an interrupt if a node goes down. In a similar case, TAF may rely on OS
timeouts to detect the failure. This takes much more time to fail over the connection than
when using FAN.

Summary

Configure client-side connect-time load balancing

Configure client-side connect-time failover
Configure server-side connect-time load balancing
Use the Load Balancing Advisory
Describe the benefits of Fast Application Notification
s a
Configure server-side callouts
) ha
Configure the server- and client-side ONS o m
y s c e
Configure Transparent Application Failover is id
un Gu br ent
@
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Upgrading and Patching Oracle RAC
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives

Describe the types of patches available

Plan for rolling patches and rolling upgrades
Install a patchset with the Oracle Universal Installer (OUI)
utility
Install a patch with the opatch utility
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Types of Patches
RDBMS patchset:
Is installed with the OUI utility

Has the naming convention: 10.2.0.4.0,11.2.0.1.0,11.2.0.3
Beginning with 11.2.0.2, patchsets are provided as full installation
packages supporting out-of-place upgrades
Merge Label Request (MLR) or one off patch:
Is installed with the opatch utility
Generally contains priority 1 fixes not in a bundle yet a
a s
Patch Set Update (PSU)
m )h
Is a cumulative patch containing recommended bug fixes o
creleased
on a quarterly schedule i s y s
i d e
r u n Gu
Is installed with the opatch utility
@ b ent
Grid PSUs often contain the Database ntoPSUSfortuthe d same version
e
im this
Critical Patch Update (CPU) s
a usec
n
Security-related tpatches
e r on e released
to quarterly
( e v ens
o n n e2012,
Copyright
v e rt rab
n E of patches
n s fe are available for Oracle products. The first type is a patchset. This
o
Different types
tra provided on top of a base release (for example, 10.2.0.1.0 or 11.1.0.2.0). A
rt of patches
isvaeset n -
Epatchsetnoincrements the fourth digit of the release number. A patchset will include updates for
Oracle Clusterware and updates for the Oracle RDBMS software. You must update Oracle
Clusterware before updating the Oracle RDBMS product. Though the reverse is not true,
Oracle Clusterware may be patched without patching the RDBMS. Patchsets are released
less frequently than other types of patches. They are cumulative and may contain hundreds of
fixes. The Oracle Universal Installer (OUI) utility is always used to install patchsets.
The second common type of patch available is known as a bundle patch (BP). A BP is a
small grouping of individual fixes, usually around 20 to 30, that is fully regression tested and
released more frequently than patchsets. Each bundle patch on the same patchset level is
cumulative for that patchset level only. The naming convention is 11.1.0.6.0 BP#1, 11.1.0.6.0
BP#2, and so on. When the patchset level is incremented, the CRS bundle patch numbering
will restart. CRS bundle patches are always applied with the opatch utility instead of the OUI
utility. BPs are binaries and do not require relinking.

Another type of patch is known as a Merge Label Request or a one off patch. These are
installed with the opatch utility and are usually priority 1 fixes that are not contained in a
bundle yet. One MLR can contain multiple fixes.
Patch Set Updates (PSU) are patches containing important updates and fixes accumulated
from the last PSU released. The purpose of these types of patches is to provide important
updates and fixes in a single, well tested package on a regular schedule. The PSUs generally
supersede older bundle patches and establish a new baseline version for easier tracking.
In January, 2005, the Critical Patch Update (CPU) became Oracles primary mechanism for
the release of security patches for all its products. CPUs are released on a quarterly basis.
The CPU schedule for the next year is posted on the Critical Patch Updates and Security
Alerts page on Oracle Technology Network (OTN). CPUs are cumulative for many Oracle
products. This means that, for these products, a CPU includes new security fixes as well as
all previously released CPU fixes for this particular platform and version combination. The
main benefit of cumulative CPUs is that they allow customers to quickly and easily catch up
to current security release level by applying only the most recent CPU.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Patch Properties
All upgrades and patchsets are installed as rolling.

All patch bundles can use the minimum down-time

procedures. Most patch bundles are rolling upgradable.
Individual patches may be rolling; check the patch.
After 11.2.0.2, patch bundles can be in-place or out-of-
place.
s a
) h a
Patch type Tool Method Rolling
o m
y s c e
Upgradable
u n is uid
Patchset OUI
b r
Out-of-place n G
Yes
t
OPatch/Enterprise ManagerntoIn-place
@ tude Most (check)
Patch bundle
c i me this S
One-off patch OPatch/Enterprise
n asManager u s e In-place Most (check)
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fe patches can have different methods and utilities for installation.
e o
Depending
rt canon their type,
trbea installed as a rolling patch, with the Oracle Universal Installer (OUI).
v
Patchsets n -
noare installed out-of-place.
EPatchsets
Patch bundles can be installed with either Enterprise Manager (EM) or OPatch. Even EM
Database Control can be used to patch Oracle Clusterware and the RDBMS software. Patch
bundles are installed in-place and most bundles can be installed as a rolling patch, but check
the README.txt file that comes with the patch, or the patch metadata.
One-off patches can only be installed in-place with EM or OPatch. Just like patch bundles,
most one-off patches can be installed as a rolling patch, if they were created to be rolling. If
the patch or patch bundle were not labeled as rolling, it cannot be used in a rolling manner.
Any one-off patch or patch bundle may use the minimum down-time method.
The use of Enterprise Manager to install patches is covered in detail in the Oracle Database 2
Day + Real Application Clusters Guide 11g Release 2.

Configuring the Software Library
Use the Provisioning page of Enterprise Manager.

Add a software library location:
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe of Enterprise Manager, you configure a software library location.
E nsfeatures
To use
e o
rcanthe patching
ra or more library locations. This will be the directory that patches will be
t have-tone
v
You n
Estored ninowhen they are transferred to the local cluster.
To navigate to the Provisioning page from the Database home page:
1. Click the Software and Support tab.
2. In the Deployment Procedure Manager section, click the Deployment and Provisioning
Software Library link.
On the Provisioning page, you will find Software Library Configuration at the bottom of the
page. The software library directory location must reference an existing directory.

Setting Up Patching
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fifethe Enterprise Manager is configured to connect to My Oracle Support
e o
Patching is simplified
tra Available patches can be downloaded and deployed through Enterprise
rt MetaLink).
v
(formerly, n -
noThe Patching Setup page can be accessed by a superuser by using the setup
EManager.
button. The patching configuration can also be supplied during installation.
To complete the patching configuration, you must supply the credentials to connect to My
Oracle Support. This assumes that your cluster has access to the Patch Search URL:
http://updates.oracle.com either directly or through a proxy. The Proxy & Connections
Settings tab allows you to configure the Internet connection.

Obtaining Oracle RAC Patches
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsand fe recommended BPs can be downloaded from the My Oracle Support
e o
The latest patchsets
t at the
rsite a
trfollowing
v
Web n - URL:
E no http://support.oracle.com/
After signing in to the Web site, click the Patches & Updates tab. For the latest patch sets
and patch bundles, click the Latest Patchsets link under Oracle Servers and Tools. You can
choose from patch sets for product bundles or patch bundles for individual products.
If you know the patchset number, you can click the Number/Name Sun CR ID link. You can
enter a single patch set number or a comma-separated list. Select your platform and click the
Search button.

s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n e
E listnofsfpatchsets
r o
For a tcomplete
t r a available for your version and platform, click the Product or
ve link.
EFamily o n-
n
After specifying Product, Release, and Platform, click the Search button. The patch search
results are displayed.

s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe patch sets, start from the Patches & Updates tab and click the
e o
To locate recommended
rt n-traPatch Advisor link. Select Oracle Clusterware from the Product pull-down
v
Recommended
o then select the release number and platform. Click the Search button for a list of
Emenu, nand
recommended patchsets.

Downloading Patches
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfethe patch you need, you can download it. Click the patch link on the
e o
Once tyou have located
r results ra Locate and click the Download link on the patch summary page, and
tpage.
v
search n -
nothe patch link in the File Download dialog box. Click the Save File button, and then
Ethen click
click OK. Specify the directory location for the file, and then click Save.

Reduced Down-Time Patching for Cluster
Environments
Patching Oracle RAC can be completed without taking the
entire cluster down.

OPatch can now apply patches in multinode, multipatch
fashion.
OPatch detects whether the database schema is at an
earlier patch level than the new patch, and runs SQL
commands to bring the schema up to the new patch level. a
ha s
OUI installs patchsets as out-of-place upgrades, reducing )
o m
the down time required for patching. sc
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E RAC s e
fcan
Patching
r t o Oracle
t r a n be completed without taking the entire cluster down. In many
ve patching
Ecases, - can be performed with zero down time. This also allows for out-of-place
non
upgrades to the cluster software and Oracle Database, reducing the planned maintenance
down time required in an Oracle RAC environment.
OPatch can now apply patches in multinode, multipatch fashion. OPatch will not start up
instances that have a nonrolling patch applied to it if other instances of the database do not
have that patch. OPatch also detects whether the database schema is at an earlier patch
level than the new patch, and it runs SQL commands to bring the schema up to the new patch
level.
You can use srvctl to shut down the Oracle software running within an Oracle home, in
preparation for patching. Oracle Grid Infrastructure patching is automated across all nodes,
and patches can be applied in a multinode, multipatch fashion.
Patchsets are now installed as out-of-place upgrades to the Grid Infrastructure software
(Oracle Clusterware and Automatic Storage Management) and Oracle Database. This
reduces the down time required for planned outages for patching.

Rolling Patches
A rolling patch allows one node at a time to be patched, while

other nodes continue to provide service. It:

Requires distinct software homes for each node
Allows different versions to coexist temporarily
May not be available for all patches
Software not shared Software shared
s a
)ha
m
co
s
ORACLE HOME Node1 GRID HOME
Node2
Node1
u n isy uide
ORACLE HOME
Node2
Local FS storage Local FS storage b nCFS

rShared t Gstorage
n t o@ tude
e is S
Rolling patches may be allowed, cim
depending on the patch. na
s se th All nodes must be
patched at the same time.
o n t o u
r t
( e ve ense
o n n e2012,
Copyright
v e rt rab
n E allows s e
fone
A rolling
r t o patch
t r a n node to be patched to the latest version, while other nodes
ve otonuse
Econtinue - the older version and provide business service. This methodology is best
n you are applying one-off patches that support rolling methodology, maintaining
suited when
high availability of your targets, so when one node is being patched, the other nodes are
available for service.
Rolling patches are enabled by using locally accessible, nonshared file system to store the
software files. Rolling patches cannot be done when the Oracle software files are stored in a
shared cluster file system in which a single copy of the software is shared among all nodes. A
single copy requires much less disk space and a single copy to patch or upgrade. However, to
patch or upgrade, the software must be stopped. Stopping the Oracle Clusterware software
also requires all databases, applications, and services that depend on Oracle Clusterware to
be stopped. This technique requires a complete outage with down time to patch or upgrade.
Note: A patchset that can be rolled for the clusterware may not be able to be rolled for the
RDBMS.

Out-of-Place Database Upgrades
You must install the software for the new Oracle Database
release before you can perform the upgrade.
The software is installed to a new Oracle Home by using OUI.

The Database Upgrade Assistant is used to finish the upgrade
process.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n ERolling n s fe
Installing
e r t o a
t r a Patchset with OUI
-
v mustoinstall
EYou n n the software for the new Oracle Database release before you can perform
the upgrade of Oracle Database. The installation procedure for the new Oracle Database 11g
Release 2 installs the Oracle software into a new Oracle home. This is referred to as an out-
of-place upgrade and is different from patch set releases for earlier releases of Oracle
Database, where the patch set was always installed in place. The new Oracle Database
software is installed using the Oracle Universal Installer. The upgrade process is completed
by the Oracle Database Upgrade Assistant.

Out-of-Place Database Upgrade with OUI
1. Upgrade Grid Infrastructure first, if necessary.

2. Follow the instructions in your Oracle OS-specific

documentation to prepare for installation of Oracle
Database software.
3. Use OUI to install the Oracle Database software.
4. Execute the Pre-Upgrade Information Tool and correct any
reported deficiencies. s a
) ha
SQL> SPOOL upgrade_info.log
SQL> @NEW_ORACLE_HOME/rdbms/admin/utlu112i.sqlsc
om
s y d e
SQL> SPOOL OFF
u i
n Gu i
r
b ent
@ tud
5. Run the root.sh script as directed e nto by SOUI.
6. Finish the upgrade process s c im DBUA.
with e t his
n na us
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsanfeOracle RAC database, then you must perform the following steps:
e o
If youtare upgrading
rUpgrade a
trGrid
E v 1. o n - Infrastructure on your cluster, if necessary.
n
2. Follow the instructions in your Oracle operating systemspecific documentation to
prepare for installation of Oracle Database software.
3. Start the Oracle Universal Installer. Select Upgrade an existing database on the Select
Installation Option page.
4. It is recommended that you run the Pre-Upgrade Information Tool before you upgrade
using DBUA, so that you can preview the types of items DBUA checks. The OUI will
launch the DBUA when the root scripts have been executed, but you can elect to run
DBUA at a later time. To use the tool, start SQL*Plus, connect to the database instance
as a user with SYSDBA privileges, and execute the utlu112i.sql script.
SQL> @NEW_ORACLE_HOME/rdbms/admin/utlu112i.sqlgfgg
5. Run the root.sh script as directed by the OUI. If you are ready, use DBUA to complete
the upgrade process.

OPatch: General Usage
To define the ORACLE_HOME or oh option on all

commands:
$ export ORACLE_HOME=/u01/app/11.2.0/grid
$ opatch command [options]
or
$ opatch command oh /u01/app/11.2.0/grid [options]
To obtain help with the OPatch syntax: s a

)ha
m
co
$ opatch command help s
u n isy uide
To check whether a patch supports a rolling b n t G
r application
(Run from the patch directory.):nto@ tude
i m e is S
$ opatch query -all |agrep s c i e th
Rolling
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n
v
Eutility s f er that the ORACLE_HOME environment variable be defined or that the
The OPatch
e
valuer - t an be passed as an argument on the command line with the oh option.
tofoORACLE_HOME
r requires
EInvgeneral,
nonORACLE_HOME refers to the home for the product to be patched. The utility
contains help for its syntax by using the help option as follows:
'opatch -help -fmw
'opatch auto -help'
'opatch apply -help'
'opatch lsinventory -help'
'opatch lspatches -help'
'opatch napply -help
'opatch nrollback -help
'opatch rollback -help'
'opatch prereq -help
In general, BPs and MLR patches can be applied in a rolling fashionthat is, one node at a
time. However, it is still important to check each patch for exceptions to this rule. To verify that
a patch supports rolling applications, unzip the downloaded patch into a directory of your
choosing and, from that directory, issue the following command:
$ORACLE_HOME/OPatch/opatch query -is_rolling_patch <patch_location>

Before Patching with OPatch
Check the current setting of the ORACLE_HOME variable.

Back up the directory being patched with an OS utility or

Oracle Secure backup.
Stage the patch to each node.
Update the PATH environment variable for the OPatch
directory.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n
The Oracle
EPatching
n s fe
utility, OPatch, verifies that the ORACLE_HOME environment variable
e o tradirectory. You should verify that the ORACLE_HOME variable is set to the
rt an actual
names
v n -
o of the product you are trying to patch.
EOraclenhome
It is best practice to back up the software directory that you are patching before performing
any patch operation. This applies to Oracle RAC, ASM, or Oracle Clusterware software
installation directories. The backup should include the Oracle Inventory directory as well.
If you manually download the patch and use OPatch to install the patch, you must stage the
patch on each node. If you use Enterprise Manager to download the patch and you selected
all the nodes in your cluster as targets for the patch, then the patch is automatically staged on
those nodes.
The opatch binary file is located in the $ORACLE_HOME/OPatch directory. You can either
specify this path when executing OPatch, or you can update the PATH environment variable to
include the OPatch directory. To change the PATH variable on Linux, use:
$ export PATH=$PATH:$ORACLE_HOME/OPatch

OPatch Automation
OPatch has automated patch application for the Oracle

Grid Infrastructure and Oracle RAC database homes.

Existing configurations are queried and the steps required
for patching each Oracle RAC database home of the same
version and the Grid home are automated.
The utility must be executed by an operating system user
with root privileges. a
h a s
OPatch must be executed on each node in the cluster if)
the Grid home or RAC home is in non-shared storage. c om
i s ys ide
One invocation of OPatch can patch the rGrid
u n home, G u one or
b RAC e t
n database
more RAC homes, or both Grid ando@ Oracle d
homes of the same Oracle releasee nt version.
S tu
im is c e th
a s
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n
v
Eutility s f er
r t o
The OPatch
t r a n automated the patch application for the Oracle Grid Infrastructure
has
ve andothe
Ehome - Oracle RAC database homes. It operates by querying existing configurations
n n the steps required for patching each Oracle RAC database home of the same
and automating
version and the GI home.
The utility must be executed by an operating system user with root privileges (usually the user
root), and it must be executed on each node in the cluster if the Grid Infrastructure home or
Oracle RAC database home is in non-shared storage. The utility should not be run in parallel
on the cluster nodes.
Depending on the command line options specified, one invocation of OPatch can patch the GI
home, one or more Oracle RAC database homes, or both GI and Oracle RAC database
homes of the same Oracle release version. You can also roll back the patch with the same
selectivity.

OPatch Automation Examples
To patch Grid home and all Oracle RAC database homes

of the same version:

# opatch auto <UNZIPPED_PATCH_LOCATION> -ocmrf
<ocm_response_file>
To patch only the GI home:

# opatch auto <UNZIPPED_PATCH_LOCATION> -oh <GI_HOME> -
ocmrf <ocm_response_file> s a
)ha
To patch one or more Oracle RAC database homes: co m
s
# opatch auto <UNZIPPED_PATCH_LOCATION> -ohnis
u
y
u ide
<oracle_home1_path>, <oracle_home2_path> b r-ocmrf n t G
<ocm_response_file>
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eutilitynwill
s e
fprompt
r t o
The OPatch
t r a for your Oracle Configuration Manager (OCM) response file
ve it isorun.
Ewhen nin- You should enter a complete path of OCM response file if you already have
creatednthis your environment. If you do not have the OCM response file (ocm.rsp), then
you should run the emocmrsp command to create it. As the software home owner, execute:
$ <ORACLE_HOME>/OPatch/ocm/bin/emocmrsp
Before executing opatch, add the directory containing opatch to your path:
# export PATH=$PATH:<GI_HOME>/OPatch
To patch GI home and all Oracle RAC database homes of the same version:
# opatch auto <UNZIPPED_PATCH_LOCATION> -ocmrf <ocm_response_file>
To patch only the GI home:
# opatch auto <UNZIPPED_PATCH_LOCATION> -oh <GI_HOME> -ocmrf
<ocm_response_file>
To patch one or more Oracle RAC database homes:
# opatch auto <UNZIPPED_PATCH_LOCATION> -oh <oracle_home1_path>,
<oracle_home2_path> -ocmrf <ocm_response_file>

To roll back the patch from the GI home and each Oracle RAC database home:
# opatch auto <UNZIPPED_PATCH_LOCATION> -rollback -ocmrf
<ocm_response_file>
To roll back the patch from the GI home:
# opatch auto <UNZIPPED_PATCH_LOCATION> -oh <path to GI home> -
rollback -ocmrf <ocm_response_file>
To roll back the patch from the Oracle RAC database home:
# opatch auto <UNZIPPED_PATCH_LOCATION> -oh <path to RAC database
home> -rollback -ocmrf <ocm_response_file>
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Quiz
Which tools can be used to install a patchset?

a. Oracle Universal Installer

b. OPatch
c. Enterprise Manager Database Console
d. Database Configuration Assistant
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o a, c
rt n-tra
v
EIn
no11g Release 2, the Oracle Universal Installer or Enterprise Manager Database
Oracle
Console can be used.

Summary

Describe the types of patches available

Plan for rolling patches and rolling upgrades
Install a patchset with the Oracle Universal Installer (OUI)
utility
Install a patch with the opatch utility
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Lesson 10 Practice Overview
The practices for this lesson covers using OPatch to patch your
Oracle Homes from 11.2.0.3 to 11.2.0.3.1.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non
Oracle RAC One Node
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives

Perform an online database migration

Add an Oracle RAC One Node Database to an existing
Cluster
Convert an Oracle RAC One Node database to a RAC
database
Use DBCA to convert a single instance database to a RACas a
h
One Node database m) co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Verifying an Existing RAC One Node Database
[oracle@host01 ~]$ srvctl config database -d orcl
Database name: orcl

Oracle user: oracle
Domain:
Start options: open
Database role: PRIMARY
Server pools: orcl s a
Database instances:
) ha
o m
Mount point paths: y s c e
Services: SERV1
u n is uid
Type: RACOneNode b r n t G
Online relocation timeout: 30
n t
Information specific to o@ tude
Instance name prefix: orcl
i me this S
RAC One Node
c
Candidate servers: host01,host02
n as use
ton e to
e r
v ens
( e
o n n e2012,
Copyright
v e rt rab
Executingn E nsfeconfig database command displays Oracle RAC One Node
the srvctl
e o tra
rt configuration
v
database n - data. The data in the output specific to Oracle RAC One Node
EincludesnoType, Online relocation timeout, and candidate servers. As you can see in the
example in the slide, the RAC One Node database orcl can run on host01 or host02 and the
online relocation timeout value is 30 minutes.
Executing the srvctl config database command without the d option returns a list of
all databases that are registered with Oracle Clusterware.

Oracle RAC One Node Online Migration
Oracle RAC One Node allows the online relocation of the

database from one server to another.

The migration period can be customized up to 12 hours.
Use the srvctl relocate database command to
initiate relocation of an Oracle RAC One Node database:
srvctl relocate database d db_unique_name {[-n target]
[-w timeout_value] | -a [-r]} [-v]
s a
-d <db_unique_name> Unique name of database to relocate
)ha
-n <target> Target node to which to relocate database m
co
s
isy uide
-w <timeout> Online relocation timeout in minutes
-a Abort failed online relocation
u n
r nt G
b
o@ tude
-r Remove target node of failed online relocation request from the
candidate server list of administrator-managed RAC One Node
n t
database
c i me this S
-v Verbose output
n as use
ton e to
-h Print usage
e r
v ens
( e
o n n e2012,
Copyright
v e rt rab
n EOne Noden s feallows the online relocation of an Oracle RAC One Node database
Oracle
e oRAC
t server
rone trato another, which provides increased availability for applications based on an
v
from
o n -
EOraclenDatabase. The migration period can be customized up to 12 hours. You can now move
a database for workload balancing as well as for performing planned maintenance on the
server, on the operating system, or when applying patches to the Oracle software in a rolling
fashion.
Only during a planned online database relocation is a second instance of an Oracle RAC One
Node database created, so that any database sessions can continue while the database is
relocated to a new node.
If your Oracle RAC One Node database is administrator managed, then the target node to
which you want to relocate the database must be in the candidate list or in the Free server
pool. If the target node is in the Free server pool, then the node is added to the candidate list.
When you relocate a database instance to a target node that is not currently in the candidate
server list for the database, you must copy the password file, if configured, to the target node.
Oracle Corporation recommends using OS authentication, instead, or using Oracle
Clusterware to start and stop the database, and defining users in the data dictionary for other
management.

Online Migration Considerations
To migrate an instance without client interruption, the

application must use client connections that are integrated

with Fast Application Notification (FAN).
You can also use connections with Transparent
Application Failover (TAF) enabled.
If FAN or TAF is not used, any in-flight transactions will be
allowed to complete within the timeout value constraint. a
a s
If the timeout is exceeded, clients will receive an ORA-3113
m )h
error when the session is terminated due to the shutdown
s co of
the instance.
u n isy uide
If the shutdown of the original instance b r longer
takes n t G than
the timeout value, the instance eisnaborted. to@ Stude
c m threcovery
iperform is
The new instance will then
n s
a us e to clean up any
transactions thatto were
n aborted
e r e to due to the shutdown.
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fewithout client interruption, the application connections must use client
e o
To migrate an
rt n-that instance
traare integrated with Fast Application Notification (FAN). For information on
v
connections
no please read the Oracle Real Application Clusters Administration and Deployment
Eusing FAN,
Guide. You can also use connections with Transparent Application Failover (TAF) enabled
(When using TAF, you should always enable FAN). If you do not use FAN or TAF, any in-
flight transactions will be allowed to complete as long as they complete within the timeout
period entered, after which the clients will receive an ORA-3113 error when their session is
terminated due to the shutdown of the Oracle RAC One Node instance (ORA-3113 End of
Line on communication channel). Because the new instance will be running, the client can
immediately log in again.
If the shutdown of the original instance takes longer than the set timeout, this database
instance is aborted. The new instance will then perform recovery to clean up any transactions
that were aborted due to the shutdown.

Performing an Online Migration
[oracle@host01 ~]$ srvctl relocate database -d orcl \

-n host02 -w 15 -v
Configuration updated to two instances

Instance orcl_2 started
Services relocated
Waiting for 15 minutes for instance orcl_1 to stop.....
s a
Instance orcl_1 stopped
) ha
Configuration updated to one instance om c e
y s
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Execute n
the srvctl
e
E nsfrelocate database command specifying the database to be
r t o t r a
ve othe
Emigrated, -host to which it will be migrated to, and optionally a timeout value used to allow
n nwith active transactions to finish. If no value is specified, the default is 30
connections
minutes. If you retrieve the status of the database before the migration starts, you should see
something like this:
[oracle@host01 ~]$ srvctl status database -d orcl
Instance orcl_2 is running on node host01
Online relocation: INACTIVE
During the relocation:
Online relocation: ACTIVE
Source instance: orcl_1 on host01
Destination instance: orcl_2 on host02
After the relocation is complete:
Online relocation: INACTIVE

Online Migration Illustration
Client connections
Server 1 Server 2 Server 3
DB-A DB-B DB-C DB-D DB-E
s a
)ha
m
co
Shared s
storage
u n isy uide
b r nt G
n t o@ tude
Single
c i me this S
n as use
cluster
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe in the slide, you have five single-instance Oracle RAC One Node
E configuration
e o
In thetexample
tra in a cluster of three servers. Server 1 is hosting Oracle RAC One Node
r nrunning
v
databases -
no DB-A (database A) and DB-B (database B), server 2 is hosting database DB-C
Edatabases
(database C) and server 3 is hosting databases DB-D (database D) and DB-E (database E).
Each server runs one OS. In servers 1 and 3, multiple databases are consolidated onto a
single OS. This deployment itself provides many consolidation benefits. However, online
database relocation, a unique feature of Oracle RAC One Node that provides live migration of
databases across nodes in the cluster, enables many additional benefits.

Client connections
DB-A DB-B DB-B DB-C DB-D DB-E
s a
)ha
m
co
Shared s
storage
u n isy uide
b r nt G
n t o@ tude
Single
c i me this S
n as use
cluster
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Onlineto n
database n fe allows an online migration of a database from one server to
E relocation
s
e r server.- a database relocation leverages the ability of Oracle Real Application
trOnline
v
another n
EClustersnoto simultaneously run multiple instances servicing a single database. In the figure in
the slide, the database B RAC One Node database on server 1 is migrated to server 2. Oracle
RAC One Node starts up a second DB-B instance on server 2, and for a short period of time
runs in an active-active configuration.
As connections complete their transactions on server 1, they are migrated to the instance on
server 2.

Client connections
s a
)ha
m
co
Shared s
storage
u n isy uide
b r nt G
n t o@ tude
Single
c i me this S
n as use
cluster
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe have migrated to the database B instance on server 2, the instance
e o
Once tall the connections
ra down and the migration is complete. To sum up, you have migrated
r 1nis-tshut
v
on server
noB from server 1 to server 2 while the database was online and performing work. To
Edatabase
extend the example, you could initiate an operating system upgrade or patch including a
reboot of server 1 by simply migrating database A (DB-A).

Online Maintenance: Rolling Patches
Client connections
Patch
DB-A DB-B DB-C DB-D DB-E

Database
binaries s a
) ha
o m
Shared
y s c e
storage
u n is uid
b r n t G
n t o@ tude
Single
c i me this S
n as use
cluster
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe
E administrators
RAC One
r t o Node
t r a can online migrate the databases off the server to a spare
ve and
Eserver, - perform all types of maintenance on the idle server, including hardware
then
non OS upgrades, OS patches, and database patches. With online database
maintenance,
relocation, a new database instance is created on a new server (running in a different
operating system), and work is online migrated to the new instance. Thus, the old operating
system and database home remain on the former host server, and can be upgraded (OS) or
patched (DB).
Lets continue with our deployment example from the previous slide. The illustration in the
slide depicts a RAC One Node deployment after using online database relocation to move
database B from server 1 to server 2. After the migration, the database binaries that had
hosted the instance formerly running on server 1 remain available for patching.

Online Maintenance: Rolling Patches
s a
)ha
m
co
Shared s
storage
u n isy uide
b r nt G
n t o@ tude
Single
c i me this S
n as use
cluster
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsbinariesfe on server 1 have been patched, database B can be migrated via
e o
Once tthe database
r databasetrarelocation back to server 1. Because online database relocation supports
v
online n -
nobetween instances at different patch levels, the operation is completely online and
Emigration
requires no disruption to end users. In this case, the online database migration migrates the
connections back to server 1, then the instance on server 2 is shut down, completing the
migration. Similarly, online database relocation can be used to move all databases off a node
in preparation for an online operating system upgrade.

Adding an Oracle RAC One Node Database to an
Existing Cluster
Use the srvctl add database command to add an
Oracle RAC One Node database to an existing cluster.

srvctl add database -c RACONENODE [-e server_list]
[-i instance_name] [-w timeout_value]
When adding an administrator-managed Oracle RAC One

Node database, you can optionally supply an instance
prefix with the -i instance_name option of the srvctl s a
add database command. ) ha
c m
ofor
Each service is configured by using the same value
ys idethe
i s
SERVER_POOLS attribute as the underlyingrundatabase.
Gu
b nt
When you add services to an Oracle
n t o@RACtuOnede Node
database, srvctl configures methose i S using the value
services
s
c i h
s
of the SERVER_POOLSaattribute. et n o us
r t o n t
v e s e
n (e 2012,
r t o n le
e a b
n E v
s f er
e r to Oracle
Converting
- t r an RAC One Node to Oracle RAC
v the srvctl
EUse non add database command to add an Oracle RAC One Node database to
an existing cluster. For example:
srvctl add database -c RACONENODE [-e server_list] [-i
instance_name] [-w timeout_value]
Use the -e option and the -i option when adding an administrator-managed Oracle RAC
One Node database.
Each service on an Oracle RAC One Node database is configured by using the same value
for the SERVER_POOLS attribute as the underlying database. When you add services to an
Oracle RAC One Node database, srvctl does not accept any placement information, but
instead configures those services using the value of the SERVER_POOLS attribute.
When adding an administrator-managed Oracle RAC One Node database, you can optionally
supply an instance prefix with the -i instance_name option of the srvctl add
database command. The name of the instance will then be instance_name_1. If you do
not specify an instance prefix, then the first 12 characters of the unique name of the database
becomes the prefix. The instance name changes to instance_name_2 during an online
database relocation and reverts to instance_name_1 during a subsequent online database
relocation. The same instance name is used on failover.

Converting a RAC One Node Database to RAC
To convert a RAC One Node database to RAC:

1. Execute the srvctl convert database command.

srvctl convert database -d <db_unique_name> -c RAC
[-n <node_1>]
2. Create server pools for each service that the database

has, in addition to the database server pool.
3. Add the instances on the remaining nodes with the s a
) ha
srvctl add instance command. om
y s c e
srvctl add instance -d <db_unique_name> -i n
u is uid
instance_name
n <node_2> b r n t G
t o@ -i u e
dinstance_name
srvctl add instance -d <db_unique_name> n t
n <node_n> c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe RAC One Node database to an Oracle RAC database by logging
E annsOracle
You can
e o convert
t Oracle
rthe traRAC One Node database owner and executing the srvctl convert
invas n -
Edatabaseno command.
After you run the command, you must create server pools for each service that the database
has, in addition to the database server pool. The values for SERVER_NAMES of the service
server pools must be set to the node that you converted from an Oracle RAC One Node to an
Oracle RAC node.
Converting an administrator-managed Oracle RAC One Node database to an Oracle RAC
database sets all database services so that the single instance is preferred. After you convert
the database, you can add instances by running the srvctl add instance command.
Converting a policy-managed Oracle RAC One Node database to an Oracle RAC database
sets all database services to UNIFORM cardinality. It also results in reusing the server pool in
which the database currently runs. The conversion reconfigures the database to run on all of
the nodes in the server pool. The command does not start any additional instances but
running the srvctl start database command starts the database on all of the nodes in
the server pool.

Converting a RAC One Node Database to RAC
[oracle@host01 ~]$ srvctl convert database -d orcl -c RAC -n host01
[oracle@host01 ~]$ srvctl add instance -d orcl -i -orcl_2 n host02
[oracle@host01 ~]$ srvctl start instance -d orcl -i orcl_2

Database name: orcl
Oracle user: oracle
Domain:
Start options: open
s a
)ha
Database role: PRIMARY m
co
s
isy uide
Server pools: orcl
u n
r nt G
Database instances: orcl_1,orcl_2 b
n t o@ tude
Services: SERV1
c i me this S
as use
Type: RAC
Database is administrator managed n
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n e
E in nthesfslide,
r o
In thetexample
t r a the RAC One Node database ORCL is converted to an Oracle
ve database
ERAC - using the srvctl convert database command. The cluster consists of
nonhost01 and host02. After the srvctl convert database command has
two nodes,
finished executing, the second instance, orcl_2 is added to host02 using the srvctl add
instance command as illustrated in the example in the slide.
Once the instance has been added to the second node, it can be started using the srvctl
start instance command. Use the srvctl config database command to verify that
the database conversion and instance addition was successful.

Converting a Single Instance Database to RAC
One Node
Use DBCA to convert from single-instance Oracle
databases to Oracle RAC One Node.

Before you use DBCA to convert a single-instance
database to an Oracle RAC One Node database, ensure
that your system meets the following conditions:
It is a supported hardware and operating system software
configuration. a
It has shared storage: either Oracle Cluster File System )orha
s
Oracle ASM is available and accessible from all nodes.
c om
Your applications have no design characteristics i s ysthat ide
r u n Gu
preclude their use in a clustered environment.
b nt
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n EDBCAnstofeconvert from single-instance Oracle databases to Oracle RAC One
You can
e o use
rtDBCA tra
v
Node. n -automates the configuration of the control file attributes, creates the undo
no and the redo logs, and makes the initialization parameter file entries for cluster-
Etablespaces
enabled environments. DBCA also configures Oracle Net Services, Oracle Clusterware
resources, and the configuration for Oracle database management for use by Oracle
Enterprise Manager or the SRVCTL utility.
Before you use DBCA to convert a single-instance database to an Oracle RAC One Node
database, ensure that your system meets the following conditions:
It is a supported hardware and operating system software configuration.
It has shared storage: either Oracle Cluster File System or Oracle ASM is available and
accessible from all nodes. On Linux on POWER systems, ensure that GPFS is available
and accessible from all nodes.
Your applications have no design characteristics that preclude their use in a clustered
environment.
Oracle strongly recommends that you use the Oracle Universal Installer to perform an Oracle
Database 11g release 2 installation that sets up the Oracle home and inventory in an identical
location on each of the selected nodes in your cluster.

Converting a RAC Database to RAC One Node
When converting a RAC database to RAC One Node, first

ensure that the RAC database has only one instance.

If the RAC database is admin-managed, change the
configuration of all services to set the preferred instance to
the one you want to keep as an RAC One Node database.
If a service had a PRECONNECT TAF policy, then the policy
must be updated to BASIC or NONE before conversion.
a
If the RAC database is policy managed, then change thehas
m)
configuration of all services so they use the same cserver
o

pool before you convert the RAC database. isys ide
Use the srvctl convert databasebrcommand un nt Guto
convert a RAC database to RAC n t
Oneo@Node:
t u de
e S s
srvctl convert database -d s c im thi
db_unique_name -c RACONENODE
a us e
nn
[-i instance_name -w timeout]
r to e to
e e
v ens
(
n e2012,
o n
rt rab l
v e
n E annsOracle fe RAC database with one instance to an Oracle RAC One Node
You can
e o convert
rt using trathe srvctl convert database command, as follows:
database
v n -
Esrvctlnoconvert database -d db_unique_name -c RACONENODE [-i
instance_name -w timeout]
Prior to converting an Oracle RAC database to an Oracle RAC One Node database, you must
first ensure that the Oracle RAC database has only one instance.
If the Oracle RAC database is administrator managed, then you must change the
configuration of all services to set the preferred instance to the instance that you want to keep
as an Oracle RAC One Node database after conversion. If any service had a PRECONNECT
TAF policy, then its TAF policy must be updated to BASIC or NONE before starting the
conversion process. These services must no longer have any available instance.
If the Oracle RAC database is policy managed, then you must change the configuration of all
services so that they all use the same server pool before you convert the Oracle RAC
database to an Oracle RAC One Node database.

Quiz
The srvctl add database command is used to add an

Oracle RAC One Node database to an existing cluster.

a. True
b. False
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o a
rt n-tra
v
no answer is True.
E correct
The

Quiz
Which of the following conditions must be met before a single

instance database can be converted to a RAC One Node

Database? (Choose three)
a. Your environment is a supported hardware and operating
system software configuration.
b. It has shared storage: either Oracle Cluster File System or
Oracle ASM is available and accessible from all nodes. a
ha s
c. You must disable Oracle Clusterware on all nodes. )
c o m
d. Your applications have no design characteristics y s that e
i s i d
run Gu
preclude their use in a clustered environment.
b en t
o @ tud
e nt S
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E d nsfe
Answer:
e o a, b,
rt n-tra
v
E no a, b, and d are correct.
Statements

Summary

Perform an online database migration

Add an Oracle RAC One Node Database to an existing
cluster
Convert an Oracle RAC One Node database to a RAC
database
Use DBCA to convert a single instance database to a RACas a
h
One Node database. m) co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Lesson 11 Practice Overview
The practices for this lesson cover the following topics:

RAC One Node Database creation using DBCA

RAC One Node online migration
Convert a RAC One Node database to a RAC Database
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Quality of Service Management
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Lesson Objectives

The purpose of Oracle Database Quality of Service (QoS)

Management
The benefits of using Oracle Database QoS Management
The components of Oracle Database QoS Management
The operation of Oracle Database QoS Management
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

QoS Management Background
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fe releases, you could use services for workload management and
e o
In previous
rt For Oracle Database
tra a group of servers might be dedicated to data warehouse work, while
v
isolation. n -example,
Eanothernisodedicated to your sales application, a third group is used for ERP processing, and a
fourth group to a custom application. Using services, the database administrator can allocate
resources to specific workloads by manually changing the number of servers on which a
database service is allowed to run. The workloads are isolated from each other, so that
demand spikes, failures and other problems in one workload do not affect the other
workloads. The problem with this type of deployment is that each workload needs to be
separately provisioned for peak demand because resources are not shared.
You could also define services that shared resources by overlapping server allocations.
However, even with this capability, you had to manually manage the server allocations and
each service was mapped to a fixed group of servers.
Starting with Oracle Database 11g, you can use server pools to logically partition a cluster
and provide workload isolation. Server pools provide a more dynamic and business-focused
way of allocating resources because resource allocations are not dependant on which servers
are up. Rather, the server pool allocations dynamically adjust when servers enter and leave
the cluster to best meet the priorities defined in the server pool policy definitions.

QoS Management Overview
Define and Enable Classify and Measure

QoS Management Policy Set

Maintenance Policy
Weekend Policy
Server
Evaluate and Policy-Driven
Pools
After Hours Policy Analyze and
Report
Architecture
Business Hours
Policy Recommend s a
Performance
Performance Objectives
)ha
Classes Business Rankings
m
co
Server Pool Allocations
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n
Implement
asanduControl s e
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E are s e
fconsolidating
r o
Manytcompanies
t r a n and standardizing their data center computer systems. In
ve with
Eparallel - the migration of applications to the Internet has introduced the problem of
nthis,
nodemand
managing surges that cannot be fully anticipated. In this type of environment, it is
necessary to pool resources and have management tools that can detect and resolve
bottlenecks in real time. Policy-managed server pools provide a foundation for dynamic
workload management. However, they can only adjust resource allocations in response to
server availability changes.
QoS Management is an automated, policy-based workload management (WLM) system that
monitors and adjusts the environment to meet business level performance objectives. Based
on resource availability and workload demands, QoS Management identifies resource
bottlenecks and provides recommendations for how to relieve them. It can make
recommendations for the system administrator to move a server from one server pool to
another, or to adjust access to CPU resources using the Database Resource Manager, in
order to satisfy the current performance objectives. Using QoS Management enables the
administrator to ensure the following:
When sufficient resources are available to meet the demand, business level
performance objectives for each workload are met, even if the workloads change.
When sufficient resources are not available to meet all demands, QoS Management
attempts to satisfy more critical business objectives at the expense of less critical ones.

QoS Management and
Exadata Database Machine
In its initial form, QoS Management is a feature of the Oracle
Database product family:

Introduced in Oracle Database 11g release 2
Associated with Oracle RAC software
Released exclusively on
Exadata Database Machine
Focused on environments supporting s a
) ha
multiple OLTP workloads QoS Management Policy m Set
o
cPolicy e
Not Exadata-specific technology y s
Maintenance Policy
Pools n
Server isAfter HoursuPolicy
Weekend
Policy-Driven
u id
The first step along the road r
bArchitecture
n G
t Policy
Business Hours
@ e
towards a broader solution tud
nto Performance
me this S
Classes Business Rankings
c i Server Pool Allocations
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfofe QoS Management is as a feature of the Oracle Database product
e o
The initial incarnation
tra with Oracle Real Application Clusters (RAC) software. It was first
rtin association
v
family n -
E no
introduced in Oracle Database 11g release 2.
The initial set of features and benefits associated with QoS Management are exclusively
available to Exadata Database Machine customers and are best suited to customers using
Database Machine predominantly as a consolidation platform for OLTP applications.
QoS Management software can operate on non-Exadata environments where Oracle
Database 11g release 2 is available. Commencing with version 11.2.0.3, a subset of QoS
Management functionality has been released that enables non-Exadata users to monitor
performance classes, but not to generate and implement changes in response to the currently
observed workload.
In its current form, QoS Management provides a powerful database-focused capability that
represents the first step along the road towards a broader workload management solution.

QoS Management Focus
Code Development
Resource Capacity Planning

Use Configuration/Deployment
Resource Quality of Service

Management
Wait s a
) ha
o m
y s c e
u n is uid
Application b r n t G
Performance n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nmonitors
s fe the performance of each work request on a target system. By
e o
QoS Management
tra the two components of performance, resource use and wait,
rt nmeasuring
v
accurately -
no can be quickly detected and resources reallocated to relieve them, thus
Ebottlenecks
preserving or restoring service levels. Changing or improving the execution time generally
requires application source code changes. QoS Management, therefore, only observes and
manages wait times.
QoS Management bases its decisions on observations of how long work requests spend
waiting for resources. Examples of resources that work requests might wait for include
hardware resources, such as CPU cycles, disk I/O queues, and global cache blocks.
Other waits can occur within the database, such as latches, locks, pins, and so on. While
these database waits are accounted for by QoS Management, they are not broken down by
type or managed. Minimizing unmanaged waits requires changes that QoS Management
cannot perform, such as application code changes and database schema optimizations for
example. QoS Management is still beneficial in these cases, because the measurement and
notification of unmanaged waits can be used as a tool to measure the effect of application
optimization activities.

QoS Management Benefits
Determines where additional resources are needed

Determines whether additional hardware can be added to

maintain acceptable performance
Reduces the number of critical performance outages
Reduces the time to resolve performance objective
violations
Improves system stability as the workload changes s a
) ha
Helps to ensure that SLAs are met o m
y s c e
Facilitates effective sharing of hardware resources
is id
r u n Gu
@ b ent
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsoffeQoS Management include:
Sometof
e o the benefits
tra and measuring database work, QoS Management can help
rBy categorizing
v n -
no
E administrators determine where additional resources are needed.
QoS Management is Oracle RACaware, and it uses this fundamental understanding to
determine if additional hardware can be added to maintain acceptable performance.
QoS Management helps reduce the number of critical performance outages. By
reallocating runtime resources to the busiest business-critical applications, those
applications are less likely to suffer from a performance outage.
QoS Management reduces the time needed to resolve performance objective violations.
Rather than requiring administrators to understand and respond to changes in
performance, much of the work can be automated. Administrators are provided with a
simple interface to review and implement the recommended changes.
Performance stresses can often lead to system instability. By moving resources to
where they are most needed, QoS Management reduces the chance that systems will
suffer from performance stress and related instability

QoS Management allows the administrator to define performance objectives that help to
ensure Service Level Agreements (SLAs) are being met. Once the objectives are
defined, QoS Management tracks performance and recommends changes if the SLAs
are not being met.
As resource needs change, QoS Management can reallocate hardware resources to
ensure that applications make more effective use of those resources. Resources can be
removed from applications that no longer require them, and added to an application that
is suffering from performance stress.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

QoS Management Functional Overview
QoS Management works with Oracle RAC and Oracle

Clusterware to:
Manage database server CPU resources by evaluating

CPU wait times to identify workloads that are not meeting
performance objectives
QoS Management can recommend:
Adjustments to the size of server pools
Alterations to consumer group mappings s a

ha
Adjustments to the CPU resources allocated to different m)

o
database instances within a server pool sc
sy de
Manage memory pressure due to number of
u n isessions
u i or

br ent G
runaway workloads o @ d nt tu
S from being
QoS Management restricts inew
c me sessions
t h i s
established on serversn s aresesuffering
athat from memory
u
stress. rton ve ense
to
( e
o n n e2012,
Copyright
v e rt rab
n E nworks s fe with Oracle RAC, Oracle Clusterware, and Cluster Health Monitor
e o
QoS Management
tra database resources to meet service levels and manage memory pressure
rt to manage
v
(CHM) n -
no servers.
Efor managed
Typically, database services are used to group related work requests and for measuring and
managing database work. For example, a user-initiated query against the database might use
a different service than a report generation application. To manage the resources used by a
service, some services may be deployed on several Oracle RAC instances concurrently,
while others may be deployed on only one instance. In an Oracle RAC database, QoS
Management monitors the nodes on which user-defined database services are offered.
Services are created in a specific server pool and the service runs on all servers in the server
pool. If a singleton service is required because the application cannot effectively scale across
multiple RAC servers, the service can be hosted in a server pool with a maximum size of one.
QoS Management periodically evaluates database server CPU wait times to identify
workloads that are not meeting performance objectives. If needed, QoS Management
provides recommendations for adjusting the size of the server pools or alterations to
Database Resource Manager (DBRM) consumer group mappings. Starting with Oracle
Database release 11.2.0.3, QoS Management also supports moving CPUs between
databases within the same server pool.

DBRM is an example of a resource allocation mechanism; it can allocate CPU shares among
a collection of resource-consumer groups based on a resource plan specified by an
administrator. A resource plan allocates the percentage of opportunities to run on the CPU.
QoS Management does not adjust DBRM plans; it activates a shared multi-level resource
plan and then, when implementing a recommendation, it moves workloads to specific
resource-consumer groups to meet performance objectives for all the different workloads.
Enterprise database servers can run out of available memory due to too many open sessions
or runaway workloads. Running out of memory can result in failed transactions or, in extreme
cases, a reboot of the server and loss of valuable resources for your applications. QoS
Management eases memory pressure by temporarily shutting down the services for database
instances on a server suffering from memory stress. This causes new sessions to be directed
to lighter loaded servers. Rerouting new sessions protects the existing workloads and the
availability of the memory-stressed server.
When QoS Management is enabled and managing an Oracle Clusterware server pool, it
receives a metrics stream from Cluster Health Monitor that provides real-time information
s a
ha
about memory resources for a server, including the amount of available memory, the amount
)
o m
of memory currently in use, and the amount of memory swapped to disk for each server. If
s c e
QoS Management determines that a node is under memory stress, the Oracle Clusterware
y
n is uid
managed database services are stopped on that node preventing new connections from being
u
b r n t G
created. After the memory stress is relieved, the services are restarted automatically and the
t o@ tude
listener can send new connections to the server. The memory pressure can be relieved in
n
c i me this S
several ways (for example, by closing existing sessions or by user intervention).
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

QoS Management Policy Sets
QoS Management Policy Set

Maintenance Policy
Weekend Policy
Server
Pools After Hours Policy
Business Hours Policy a
ha s
m )
Performance Objectives o
c e
Performance y s
u n is uid
Classes Business
b r Rankings
t G
n
n t o@ tude
meServer i s S Allocations
Pool
c i h
n as use t
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E ninsQoS fe Management is the policy set. A policy set allows you to specify
A central
e t o concept
rresources, traperformance classes (workloads), and a collection of performance policies
v
your n -
no the performance objective for each performance class and sets constraints for
Ethat specify
resource availability. QoS Management uses a system-wide policy set that defines
performance objectives based upon the classes of work and the availability of resources.
Specific performance policies can be enabled based upon a calendar schedule, maintenance
windows, events, and so on. Only one performance policy can be in effect at any time.
To maintain the current performance objectives, QoS Management makes resource
reallocation recommendations and predicts their effect. The recommendations can be easily
implemented with a single button click.
A policy set consists of the following:
The server pools that are being managed by QoS Management
Performance classes, which are work requests with similar performance objectives
Performance policies, which describe how resources should be allocated to the
performance classes by using performance objectives and server pool directive
overrides. Within a performance policy, performance objectives are ranked based on
business importance, which enables QoS Management to focus on specific objectives
when the policy is active.

Server Pools
Sales Sales ERP HR Batch

App Cart App App App
ERP HR Batch
Service Service Service
SALES Service
s a
)h a
m
Online Server Pool (SP) BackOffice SP Batchy s
SP
co Free
e SP
Min=3, Max=4, Min=2, Max=2, i s i d
n Max=1,Gu Min=0,
Min=1,
Importance=30 Importance=20 r
b ent u
Importance=10 Imp=0
@ tud
SALES Database APPS
e nto Database S
s c im this
Oracle
n a Clusterware
u s e
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe division of a cluster. Server pools facilitate workload isolation within
E is a nlogical
e o
A server pool
rt while a
trmaintaining
avcluster n - agility and allowing users to derive other benefits associated with
no Administrators can define server pools, which are typically associated with
Econsolidation.
different applications and workloads. An example is illustrated in the slide. QoS Management
can assist in managing the size of each server pool and also by managing the allocation of
resources within a server pool.
When Oracle Grid Infrastructure is first installed, a default server pool, called the Free pool, is
created. All servers are initially placed in this server pool. Specific server pools can then be
created for each of the workloads that needs to be managed. When a new server pool is
created, the servers assigned to that server pool are automatically moved out of the Free pool
and placed into the newly created server pool.
After a server pool is created, a database can be configured to run on the server pool, and
cluster-managed services can be established for applications to connect to the database.
For an Oracle RAC database to take advantage of the flexibility of server pools, the database
must be created using the policy-managed deployment option, which places the database in
one or more server pools.

A key attribute of policy-based management is the allocation of resources to server pools
based on cardinality and importance.
When the cluster starts or when servers are added, all the server pools are filled to their
minimum levels in order of importance. After the minimums are met, server pools continue to
be filled to their maximums in order of importance. If there are any left-over servers, they are
allocated to the Free pool.
If servers leave the cluster for any reason, a server reallocation may take place. If there are
servers in the Free pool and another server pool falls below its maximum value, a free server
is allocated to the affected server pool. If there are no free servers, then server reallocation
takes place only if a server pool falls below its minimum level. If that occurs, a server will be
sourced from one of the following locations in the following order:
1. The server pool with the lowest importance that has more than its minimum number of
servers
2. The server pool with the lowest importance that has at least one server and has lower
importance than the affected server pool s a
) ha
Using these mechanisms, server pools can maintain an optimal level of resources based on
o m
the current number of servers that are available.
y s c e
u n is uid
Consider the example shown in the slide. If one of the servers in the Online server pool failed,
b r n t G
the server currently residing in the Free server pool would automatically move to the Online
server pool.
n t o@ tude
Now, if one of the servers from the BackOffice server pool failed, there would be no servers to
i me this S
allocate from the Free server pool. In this case, the server currently servicing the Batch server
c
n as use
pool would be dynamically reallocated to the BackOffice server pool, because the failure
e r ton e to
would cause the BackOffice server pool to fall below its minimum and it has a higher
( e v ens
importance than the Batch server.
o n n e lic
If one node is later returned to the cluster, it will be allocated to the Batch pool in order to
e r t a b l
satisfy the minimum for that server pool.
Ev nsfer
Any additional nodes added to the cluster after this point will be added to the Free pool,
n
e r to -tra
because all the other pools are filled to their maximum level.
Ev non

Performance Classes
A performance class is a group of work requests whose

service level needs to be managed.

Work requests are defined by performance classifiers
containing the database service name and optional
session parameters.
An initial set of performance classifiers is automatically
discovered and created from cluster-managed services. a
ha s
Performance objectives are defined on performance )
classes. c om
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eclasses s e
fare
r t o
Performance
t r a n used to categorize workloads with similar performance
ve on- A set of classification rules are evaluated against work requests when they
Erequirements.
n
arrive at the edge of the system. These rules allow value matching against attributes of the
work request; when there is a match between the type of work request and the criteria for
inclusion in a performance class, the work request is classified into that performance class.
This classification of work requests applies the user-defined name, or tag, that identifies the
performance class (PC) to which the work request belongs. All work requests that are
grouped into a particular PC have the same performance objectives. In effect, the tag
connects the work request to the performance objective that applies to it. Tags are carried
along with each work request so that every component of the system can take measurements
and provide data to QoS Management for evaluation against the applicable performance
objectives.
QoS Management supports user-defined combinations of connection parameters called
classifiers to map performance classes to the actual workloads running in the database.

These connection parameters fall into two general classes and can be combined to create
fine-grained Boolean expressions:
Configuration Parameters: The supported configuration parameters are
SERVICE_NAME and USERNAME. Each classifier in a performance class must include
one or more cluster-managed database services. Additional granularity can be achieved
by identifying the Oracle Database user that is making the connection from either a
client or the middle tier. The advantage of using these classifiers is that they do not
require application code changes to define performance classes.

Application Parameters: The supported application parameters are MODULE,
ACTION, and PROGRAM. These are optional parameters set by the application as
follows:
- OCI: Use OCI_ATTR_MODULE and OCI_ATTR_ACTION.
- ODP.NET: Specify the ModuleName and ActionName properties on the
OracleConnection object.
s a
- JDBC: Set MODULE and ACTION in SYS_CONTEXT. a
) hplatform.
The PROGRAM parameter is set or derived differently for each database driverm and
Please consult the appropriate Oracle Database developers guide for further s codetails and
examples.
u n isy uide
To manage the workload for an application, the application code b t G
rdirectsndatabase
connections to a particular service. The service namenis t @ tuindaeclassifier, so all work
ospecified
requests that use that service are tagged as belonging
c i me ttohthe is Sperformance class created for
that application. If you want to provide more
n s
a precises econtrol over the workload generated by
various part of the application, you o n
can create u
additional
o performance classes and use
r t t
classifiers that include MODULE,
( e v e ACTION,
e n seor PROGRAM in addition to the SERVICE_NAME
or USERNAME.
o n n e lic
The performance e r t
classes b
used
a l in an environment can change over time. A common scenario
n s er
is to replaceEavsingle performance
f objective with multiple, more specific performance
r t o
objectives,
t r a
dividing n the work requests into additional performance classes. For example,
v e
Eapplication o -
ndevelopers can suggest performance classes for QoS Management to use. In
n
particular, an application developer can define a collection of database classifiers using the
MODULE and ACTION parameters, and then put them in separate performance classes so
each type of work request is managed separately.

Classification and Tagging
Each session is classified:

The classification is determined by evaluating session

parameters against performance class classifiers.
Evaluation occurs only when a session is established or
when session parameters change.
This minimizes the overhead associated with classification.
Each work request is tagged:
The tag is based on the current session classification. ) ha
sa
m
co class.
The tag connects the work request with a performance
s

u u ideto be
isy request
It enables measurements associated with thenwork
nt G
br class.
recorded against the appropriate performance
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe
E Management,
r t o
To enable QoS
t r a work requests must be classified and tagged.
ve a database
EWhen o n- session is established, the session parameters are evaluated against the
n
performance class classifiers to determine a classification. Work associated with the session
is then tagged based on the session classification until the session ends or the session
parameters change. If the session parameters change, the classification is re-evaluated. Thus
the overhead associated with classification is very small, because the classification is only
evaluated when a session is established or when session parameters change.
Tags are permanently assigned to each work request so that all the measurements
associated with the work request can be recorded against the appropriate performance class.
In effect, the tag connects the work request to a performance class and its associated
performance objective.

Performance Policies
Performance policies are named sets of performance

objectives and server pool overrides to meet business

objectives.
Performance objectives can be ranked according to their
importance.
Only one policy is active at any time.
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe
Evariousnsperformance
To manage
r t o t r a objectives, a QoS Management administrator defines one or
ve performance
Emore - policies. For example, the administrator might define a performance policy
nonbusiness hours, another for weekday non-business hours, one for weekend
for normal
operations, and another to be used during processing for the quarter-end financial closing.
Note that at any time, only one performance policy is in effect.
A performance policy has a collection of performance objectives in effect; one or more for
each application that is being managed on the system. Some performance objectives are
always more critical to the business than others, while other performance objectives might be
more critical at certain times, and less critical at other times. The ability to define multiple
performance policies inside the policy set provides QoS Management with the flexibility
required to implement different priority schemes when they are required.

Performance Class Ranks
Performance class ranks assign a relative level of

business criticality to each performance class within a

performance policy:
Highest
High
Medium
Low a
a s
Lowest
m )h
s co
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
E nsfpolicy, e
Withintoan
performance
a you can also rank each performance class. This rank assigns a
r
ve level t r
n-of business criticality to each performance objective. When there are not enough
Erelative
noavailable
resources to meet all the performance objectives for all performance classes, the
performance objectives for the more critical performance classes must be met at the expense
of the less critical ones. The available rank settings are Highest, High, Medium, Low, or
Lowest. Note that if more than one class is assigned a particular rank (for example, Medium),
classes are then ordered within that ranking alphabetically.

Performance objectives can be derived from your SLAs.

They specify:
A business requirement
The performance class to which the business requirement
applies
Average response time per database call is currently the
only performance objective type. a
a s
Response time is the total time from the time the database
m )h
receives the request to when the response leaves the co server.
Response time does not include network traffic i s ytime. ide
s
r u n Gu
@ b ent
e nto Stud
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fe objective for each performance class to specify the desired
e o
You create a performance
tra for that performance class. A performance objective specifies both a
rt n-level
v
performance
norequirement, and the work to which it applies (the performance class). For example,
Ebusiness
a performance objective might say that database work requests that use the SALES service
should have an average response time of less than 60 milliseconds.
Each performance policy includes a performance objective for each and every performance
class, unless the performance class is marked measure-only. In this release, QoS supports
only one type of performance objective, average response time.
Response time is based upon database client calls from the point that the database server
receives the request over the network until the request leaves the server. Response time
does not include the time it takes to send the information over the network to or from the
client. The response time for all database client calls in a performance class is averaged and
presented as the average response time.

Performance Satisfaction Metrics
Different performance objectives can be compared using

the Performance Satisfaction Metric (PSM).

The PSM quickly shows how the system is coping with the
objective.
100%
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me thi-100% sS
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfeobjectives are used to measure the performance of different workloads.
e t o
Different performance
rManagementtra currently supports only OLTP workloads and uses only the average
v
QoS n -
Eresponsenotime performance objective. When configuring QoS Management, you can have
very different performance objectives for each performance class. For example, one
performance objective may specify that a Checkout call should complete within 1 millisecond,
while another performance objective may specify that a Browse call should complete within 1
second. As more performance objectives are added to a system, it can be difficult to compare
them quickly.
Because of this, it is useful to have a common and consistent numeric measure indicating
how the current workload for a performance class is measuring up against its current
performance objective. This numeric measure is called the Performance Satisfaction Metric.
The Performance Satisfaction Metric is thus a normalized numeric value (between +100%
and -100%) that indicates how well a particular performance objective is being met, and which
allows QoS Management to compare the performance of the system for widely differing
performance objectives.

Server Pool Directive Overrides
Business Hours Policy

Online SP Back Office SP Batch SP
After Hours Policy
s a
)ha
Batch SP o m
Online SP Back Office SP
y s c e
End Of Quarter Policy
u n is uid
b r n t G
n t o@ tude
c i me this S
n a s se
Online SP
o
n Back Office
o uSP Batch SP
r t t
( e ve ense
o n n e2012,
Copyright
v e rt rab
n E policy s e
fcan
r t o
A performance
t r a n also include a set of server pool directive overrides. A server pool
ve ooverride
Edirective - sets the minimum server count, maximum server count, and importance
n forn a server pool when the performance policy is in effect. Server pool directive
attributes
overrides serve as constraints on the recommendations proposed by QoS Management,
because the server pool directive overrides are honored while the performance policy is
active. For example, QoS Management will never recommend moving a server out of a server
pool if doing so will leave the server pool below its minimum server count value.
Server pool directive overrides can be used to define the normal state of server pools at
different points in time. The slide illustrates an example. Under normal conditions, these
server pool settings would be expected to handle the prevailing workload. If there is a sudden
increase in the workload requests for a performance class, then the associated server pool
might require additional resources beyond what is specified in the performance policy.

Overview of Metrics
QoS Management uses a standardized set of metrics.

There are two metric types:

Performance metrics give an overview of where time is spent
in the system.
Resource metrics measure the time that work requests use a
resource or wait for a resource.
Metrics are used to identify bottlenecked resources and to a
determine the best corrective action: h a s
) m
For a performance class, the bottlenecked resource o the
cis
resource that contributes the largest average i s y time.ide
wait
s
n u ru nt G
b
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nuses s fea standardized set of metrics, which are collected by all the servers in
e o
QoS Management
rt nTheretra are two types of metrics: performance metrics and resource metrics. These
v
the system. -
o direct observation of the use and wait time incurred by work requests in each
Emetricsnenable
performance class, for each resource requested, as it traverses the servers, networks, and
storage devices that form the system.
Performance metrics are collected at the entry point to each server in the system. They give
an overview of where time is spent in the system and enables comparisons of wait times
across the system. Data is collected periodically and forwarded to a central point for analysis,
decision-making, and historical storage.
A key performance metric is response time, or the difference between the time a request
comes in and the time a response is sent out. The response time for all database calls in a
performance class is averaged and presented as the average response time. Another
important performance metric is the arrival rate of work requests. This provides a measure of
the demand associated with each performance class.

Resource metrics exist for the following resources; CPU, Storage I/O, Global Cache, and
Other (database waits). Two resource metrics are provided for each resource:
Resource usage time: Measures how much time is spent using the resource
Resource wait time: Measures the time spent waiting to get the resource
QoS Management metrics provide the information needed to systematically identify
performance class bottlenecks in the system. When a performance class is violating its
performance objective, the bottleneck for that performance class is the resource that
contributes the largest average wait time for each work request in that performance class.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

QoS Management Architecture
Data Sources QoS Management Enterprise

Server Manager
Oracle
Data Connectors and Manager

RAC JMX Client Management
11.2 JDBC
Interfaces Dashboard
XML
Policy Set
Wizard
Oracle
Clusterware Policy Engine s a
11.2 SRVM
Performance
History)ha
Oracle Business Rules
m
co&
s
Operating u n isyActions
Alerts
u ide
System b r nt G
CHM Persistence@
Manager n t o t u de
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n s fe metrics data from each database instance running in managed
E nretrieves
e o
QoS Management
tracorrelates the data by performance class every 5 seconds. The data
rt poolsn-and
v
server
Eincludesnomany metrics; for example, call arrival rate and CPU, I/O and Global Cache use, and
wait times. The data is combined with the current topology of the cluster and the health of the
servers in the Policy Engine to determine the overall performance profile of the system with
regard to the current performance objectives established by the active performance policy.
The performance evaluation occurs once a minute and results in a recommendation if there is
a performance class not meeting its objective. The recommendation specifies what resource
is bottlenecked. Specific corrective actions are included, if possible, along with the projected
impact on all performance classes in the system. The slide shows the collection of data from
various data sources by the data connectors component of QoS Management:
Oracle RAC 11.2 communicates with the data connector using JDBC.
Oracle Clusterware 11.2 communicates with the data connector using the SRVM
component of Oracle Clusterware.
The server operating system communicates with the data connector using Cluster
Health Monitor (CHM).
Enterprise Manager displays the information in a variety of ways (for example, on the
Management Dashboard, Policy Set Wizard, Performance History, and Alerts and Actions
pages).

QoS Management Recommendations
If performance objectives are not being met, QoS

Management makes a recommendation.

Each recommendation focuses on improving the highest
ranked performance class not exceeding its performance
objective.
Recommendations may include:
Changing consumer group mappings s a
Reprioritize work within existing resource boundaries. )ha
co m
Moving servers between server pools s
Reprioritize resources between server poolsuto n u ide
sy workload
imeet
r nt G

demands.
@ tudeb
Moving CPUs between databases n t owithin
i m e is S a server pool
Reprioritize CPU resources s c within e h
texisting server pool

boundaries. n a u s
r t o n t o
( e ve ense
o n n e2012,
Copyright
v e rt rab
If yourto n
business n fe periodic demand surges, then to retain performance levels for
E experiences
s
v
youre n - trayou can acquire additional hardware to be available when needed, and sit
rapplications
nonot needed. Rather than have extra servers sit idle for most of the time, you might
Eidle when
decide to use those servers to run other application workloads. However, if the servers are
busy running other applications when a demand surge hits, your main business applications
are not able to perform as expected. QoS Management helps to manage such situations.
When you implement a performance policy, QoS Management continuously monitors the
system and manages it using an iterative process. When one or more performance objectives
are not being met, each iteration seeks to improve the performance of a single performance
objective; the highest ranked performance objective that is currently not being met. When all
performance objectives are being met, QoS Management makes no further
recommendations.
The recommendations take the form of moving servers between server pools, changing
consumer group mappings, or moving CPUs between databases within a server pool.
Changing consumer group mappings may involve promoting a specific workload so that it
gets a greater share of resources, or it may involve demoting a competing workload as a way
of making additional resources available to the target performance class. In both cases,
workloads are reprioritized within existing resource boundaries.

Moving servers between server pools is another approach used by QoS Management. This
approach alters the distribution of servers to meet workload demands.
Commencing with Oracle Database release 11.2.0.3, QoS Management can also move CPU
resources between databases within the same server pool. This alters the distribution of CPU
resources between database instances using instance caging and provides additional control
for environments where multiple databases are consolidated within the same Exadata
Database Machine environment.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Implementing Recommendations
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fe is working to improve the performance of a particular performance
e o
WhentQoS Management
tra to add more of the bottleneck resource (such as CPU time) for that
r it recommends
v
class, n -
no class, or to make the bottleneck resource available more quickly to work
Eperformance
requests in the performance class.
Implementing a recommendation makes the resource less available to other performance
classes. The negative impact on the performance classes from which the resource is taken
may be significantly smaller than the positive impact on the service that is getting better
access, resulting in a net win for the system as a whole. Alternatively, the performance class
being penalized may be less business critical than the one being helped.
When generating recommendations, QoS Management evaluates the impact to the system
performance as a whole. If the improvement for one performance class is rather small, but the
negative impact on another performance class is large, then QoS Management might report
that the performance gain is too small, and not recommended. If there is more than one way
to resolve the bottleneck, then QoS Management advises the best overall recommendation
factoring in variables such as the calculated impact on all the performance classes along with
the predicted disruption and settling time associated with the action. Using Oracle Enterprise
Manager, you can view the current recommendation and the alternative recommendations.

Performance data is sent to Oracle Enterprise Manager for display on the QoS Management
Dashboard and Performance History pages. Alerts are generated to drive notifications that
one or more performance objectives are not being met or that a problem has developed that
prevents one or more server pools from being managed. As a result of these notifications the
administrator can implement the recommendation.
In this release, QoS Management does not implement the recommendations automatically. It
suggests a way of improving performance, which must then be implemented by the
administrator by clicking the Implement button. After implementing a recommendation, the

system is allowed to settle before any new recommendations are made. This is to ensure
stable data is used for further evaluations and also to prevent recommendations that result in
oscillating actions.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Quiz
Oracle Database Quality of Service Management helps to meet

performance objectives by reducing resource usage.

a. True
b. False
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o b
rt n-tra
E v
Oracle o
nDatabase of Service Management helps to meet performance objectives by managing
and reducing resource wait times, not resource usage.

Quiz
Oracle Database Quality of Service Management

recommendations can include:

a. Moving servers between server pools
b. Adding spindles to improve I/O performance
c. Changing consumer group mappings
d. Recommending partitioning strategies to ease global
cache bottlenecks s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
Answer:
e o a, c
rt n-tra
E v
Oracle o
nDatabase Quality of Service Management can identify I/O performance issues or
global cache bottlenecks, but cannot address them in this release.

Summary
In this lesson, you should have learned how to describe:

The purpose of Oracle Database Quality of Service (QoS)

Management
The benefits of using Oracle Database QoS Management
The components of Oracle Database QoS Management
The operation of Oracle Database QoS Management
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Lesson 12 Demonstrations
Configuring Quality of Service Management

Using Quality of Service Management
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Design for High Availability
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no
Objectives

Design a Maximum Availability Architecture in your

environment
Determine the best RAC and Data Guard topologies for
your environment
Configure the Data Guard Broker configuration files in a
RAC environment s a
ha
Identify successful disk I/O strategies m) co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

Causes of Unplanned Down Time
Unplanned Down time

Software failures Hardware failures Human errors Disasters
Operating system CPU Operator error Fire
Database Memory User error Flood
Middleware Power supply DBA Earthquake

s a
Application Bus System admin. Power failure
) ha
o m
Network Disk Sabotage
y s c e
Bombing
Tape u n is uid
b r n t G
Controllers
n t o@ tude
Network
c i me this S
Powernn
as use
e r to e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe in designing a highly available solution is examining and
E challenges
One of
e o
the true
ra possible causes of down time. It is important to consider causes of both
rt nall-tthe
v
addressing
no and planned down time. The diagram in the slide, which is a taxonomy of
Eunplanned
unplanned failures, classifies failures as software failures, hardware failures, human error,
and disasters. Under each category heading is a list of possible causes of failures related to
that category.
Software failures include operating system, database, middleware, application, and network
failures. A failure of any one of these components can cause a system fault.
Hardware failures include system, peripheral, network, and power failures.
Human error, which is a leading cause of failures, includes errors by an operator, user,
database administrator, or system administrator. Another type of human error that can cause
unplanned down time is sabotage.
The final category is disasters. Although infrequent, these can have extreme impacts on
enterprises, because of their prolonged effect on operations. Possible causes of disasters
include fires, floods, earthquakes, power failures, and bombings. A well-designed high-
availability solution accounts for all these factors in preventing unplanned down time.

Causes of Planned Down Time
Planned down time

Routine operations Periodic maintenance New deployments
Backups Storage maintenance HW upgrade
Performance mgmt Initialization parameters OS upgrades
Security mgmt Software patches DB upgrades

s a
Batches Schema management MidW upgrades
)h a
Operating system o m
Appcupgrades
i s y s
i d e
Middleware
r u n NetGupgrades
u
b en t
Network o @ tud
e nt S
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E timenscan febe just as disruptive to operations, especially in global enterprises
Planned
e o down
rt nusers tra in multiple time zones, up to 24 hours per day. In these cases, it is
v
that support -
noto design a system to minimize planned interruptions. As shown by the diagram in
Eimportant
the slide, causes of planned down time include routine operations, periodic maintenance, and
new deployments. Routine operations are frequent maintenance tasks that include backups,
performance management, user and security management, and batch operations.
Periodic maintenance, such as installing a patch or reconfiguring the system, is occasionally
necessary to update the database, application, operating system middleware, or network.
New deployments describe major upgrades to the hardware, operating system, database,
application, middleware, or network. It is important to consider not only the time to perform the
upgrade, but also the effect the changes may have on the overall application.

Oracles Solution to Down Time
RMAN backup/recovery
Fast-Start
RAC
Fault Recovery
Data Guard
ASM
Streams
System
failures Flashback
Unplanned
down time
Data HARD
failures
Data Guard s a
& ) ha
Streams o m
Rolling upgrades/
y s c e
System Online patching
u n is uid
changes
Dynamic provisioning b r n t G
Planned
down time n t o@ tude
Data c i me this S
changes na
s Online s e redefinition
u
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s feis primarily the result of computer failures or data failures. Planned
e o
Unplanned down time
tra due to data changes or system changes:
rttime nis-primarily
v
down
E RAC no provides optimal performance, scalability, and availability gains.
Fast-Start Fault Recovery enables you to bound the crash/recovery time. The database
self-tunes checkpoint processing to safeguard the desired recovery time objective.
ASM provides a higher level of availability using online provisioning of storage.
Flashback provides a quick resolution to human errors.
Oracle Hardware Assisted Resilient Data (HARD) is a comprehensive program
designed to prevent data corruptions before they happen.
Recovery Manager (RMAN) automates database backup and recovery.
Data Guard must be the foundation of any Oracle database disaster-recovery plan.
The increased flexibility and capability of Streams over Data Guard with SQL Apply
requires more expense and expertise to maintain an integrated high availability solution.
With online redefinition, the Oracle database supports many maintenance operations
without disrupting database operations, or users updating or accessing data.
Oracle Database continues to broaden support for dynamic reconfiguration, enabling it
to adapt to changes in demand and hardware with no disruption of service.
Oracle Database supports the application of patches to the nodes of a RAC system, as
well as database software upgrades, in a rolling fashion.
RAC and Data Guard Complementarity
Resource Cause Protection
Nodes RAC
Component
Instances failure RAC
Software failure s a
)ha
Human m
error
Data Guard
s co
Data
Environment
u n isy& uide
b r nt G
Flashback
n t o@ tude
i m e is S Data Guard
Site
a s c e th &
o n n o us Streams
t
er nse t
e v
( 2012,
n n
e r to able
n EvGuard s f er
RAC and
r t o Data
t r a n together provide the benefits of system-level, site-level, and data-level
ve onresulting
Eprotection, - in high levels of availability and disaster recovery without loss of data.
n
RAC addresses system failures by providing rapid and automatic recovery from failures,
such as node failures and instance crashes.
Data Guard addresses site failures and data protection through transactionally
consistent primary and standby databases that do not share disks, enabling recovery
from site disasters and data corruption.
Note: Unlike Data Guard using SQL Apply, Oracle Streams enables updates on the replica
and provides support for heterogeneous platforms with different database releases.
Therefore, Oracle Streams may provide the fastest approach for database upgrades and
platform migration.

Maximum Availability Architecture
Real-time query
Clients
Oracle Oracle
Application Application
Server Server
WAN Traffic
Manager
Real-time query
s a
) ha
Primary o
Secondary m
site Data Guard y s c e
site
u n is uid
b r n t G
n t o@ tude
RAC c i me this S RAC databases:
database n as use Phys&log standby
t o n t o
e v er nse
n ( 2012,
r n
to able
e
EvGuard er the basis for the database MAA solution. MAA provides the
n s f
RAC and
e
most r to -tran architecture for reducing down time for scheduled outages and
Data
comprehensive
provide
v
nondetecting, and recovering from unscheduled outages. The recommended MAA
Epreventing,
has two identical sites. The primary site contains the RAC database, and the secondary site
contains both a physical standby database and a logical standby database on RAC. Identical
site configuration is recommended to ensure that performance is not sacrificed after a failover
or switchover. Symmetric sites also enable processes and procedures to be kept the same
between sites, making operational tasks easier to maintain and execute.
The graphic illustrates identically configured sites. Each site consists of redundant
components and redundant routing mechanisms, so that requests are always serviceable
even in the event of a failure. Most outages are resolved locally. Client requests are always
routed to the site playing the production role.
After a failover or switchover operation occurs due to a serious outage, client requests are
routed to another site that assumes the production role. Each site contains a set of application
servers or mid-tier servers. The site playing the production role contains a production
database using RAC to protect from host and instance failures. The site playing the standby
role contains one standby database, and one logical standby database managed by Data
Guard. Data Guard switchover and failover functions allow the roles to be traded between
sites.

RAC and Data Guard Topologies
Symmetric configuration with RAC at all sites:

Same number of instances

Same service preferences
Asymmetric configuration with RAC at all sites:
Different number of instances
Different service preferences
Asymmetric configuration with mixture of RAC and single as a
h
instance: m) o
All sites running under Oracle Clusterware y s c e
u n is uid
b r
Some single-instance sites not running under
n t G
Oracle
Clusterware to@ de e n S tu
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe database to protect a primary database in a RAC environment.
E nasstandby
You can
e o configure
tra of combinations are supported. For example, it is possible to have your
rt alln-kinds
v
Basically,
o
Eprimaryndatabase running under RAC, and your standby database running as a single-
instance database. It is also possible to have both the primary and standby databases running
under RAC.
The slide explains the distinction between symmetric environments and asymmetric ones.
If you want to create a symmetric environment running RAC, then all databases need to have
the same number of instances and the same service preferences. As the DBA, you need to
make sure that this is the case by manually configuring them in a symmetric way.
However, if you want to benefit from the tight integration of Oracle Clusterware and Data
Guard Broker, make sure that both the primary site and the secondary site are running under
Oracle Clusterware, and that both sites have the same services defined.
Note: Beginning with Oracle Database 11g, the primary and standby systems in a Data
Guard configuration can have different CPU architectures, operating systems (for example,
Windows and Linux for physical standby database only with no EM support for this
combination), operating system binaries (32-bit and 64-bit), and Oracle database binaries (32-
bit and 64-bit).

RAC and Data Guard Architecture
Primary instance A Standby receiving instance C

ARCn
ARCn LGWR RFS
Flash
Primary recovery
database Standby area
Online redo
redo files s a
files Standby )ha
Flash m
recovery databaseco
area LGWR RFS i s ys ide
r u n Gu
b Apply e n t
@ tud
ARCn
e nto ARCn S
s c im this
Primary instance B n a usStandby e apply instance D
n
rto se t o
n (e 2012,
r t o n le
e a b
n
v
Eperfectly s f er
Although
r t o it is
t r a n possible to use a RAC to single-instance Data Guard (DG)
ve on- you also have the possibility to use a RAC-to-RAC DG configuration. In this
Econfiguration,
mode, n although multiple standby instances can receive redo from the primary database, only
one standby instance can apply the redo stream generated by the primary instances.
A RAC-to-RAC DG configuration can be set up in different ways, and the slide shows you one
possibility with a symmetric configuration where each primary instance sends its redo stream
to a corresponding standby instance using standby redo log files. It is also possible for each
primary instance to send its redo stream to only one standby instance that can also apply this
stream to the standby database. However, you can get performance benefits by using the
configuration shown in the slide. For example, assume that the redo generation rate on the
primary is too great for a single receiving instance on the standby side to handle. Suppose
further that the primary database is using the SYNC redo transport mode. If a single receiving
instance on the standby cannot keep up with the primary, then the primarys progress is going
to be throttled by the standby. If the load is spread across multiple receiving instances on the
standby, then this is less likely to occur.
If the standby can keep up with the primary, another approach is to use only one standby
instance to receive and apply the complete redo stream. For example, you can set up the
primary instances to remotely archive to the same Oracle Net service name.

You can then configure one of the standby nodes to handle that service. This instance then
both receives and applies redo from the primary. If you need to do maintenance on that node,
then you can stop the service on that node and start it on another node. This approach allows
for the primary instances to be more independent of the standby configuration because they
are not configured to send redo to a particular instance.
Note: For more information, refer to the Oracle Data Guard Concepts and Administration
guide.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Data Guard Broker (DGB) and
Oracle Clusterware (OC) Integration
OC manages intrasite high availability (HA) operations.
OC manages intrasite planned HA operations.

OC notifies when manual intervention is required.
DBA receives notification.
DBA decides to switch over or fail over using DGB.
DGB manages intersite planned HA operations.
s a
DGB takes over from OC for intersite failover, switchover,
) ha
and protection mode changes: c o m
y s e or
DMON notifies OC to stop and disable the site, i s
n Gu leaving
i dall
one instance. r
b ent u
@
o thetusite d according to
DMON notifies OC to enable and
e ntstart S
the DG site role.
s c im this
n a use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n n s fe with Oracle Clusterware. Oracle Clusterware manages individual
E integrated
DGB is
e otightly
tra unattended high availability of a given clustered database. DGB
rt ton-provide
v
instances
Emanagesnoindividual databases (clustered or otherwise) in a Data Guard configuration to
provide disaster recovery in the event that Oracle Clusterware is unable to maintain
availability of the primary database.
For example, Oracle Clusterware posts NOT_RESTARTING events for the database group and
service groups that cannot be recovered. These events are available through Enterprise
Manager, ONS, and server-side callouts. As a DBA, when you receive those events, you
might decide to repair and restart the primary site, or to invoke DGB to fail over if not using
Fast-Start Failover.
DGB and Oracle Clusterware work together to temporarily suspend service availability on the
primary database, accomplish the actual role change for both databases during which Oracle
Clusterware works with the DGB to properly restart the instances as necessary, and then to
resume service availability on the new primary database. The broker manages the underlying
Data Guard configuration and its database roles, whereas Oracle Clusterware manages
service availability that depends upon those roles. Applications that rely upon Oracle
Clusterware for managing service availability will see only a temporary suspension of service
as the role change occurs within the Data Guard configuration.

Fast-Start Failover: Overview
Fast-Start Failover implements automatic failover to a

standby database:
Triggered by failure of site, hosts, storage, data file offline
immediate, or network
Works with and supplements RAC server failover
Failover occurs in seconds (< 20 seconds).
Comparable to cluster failover a
a s
The original production site automatically rejoins the ) h
configuration after recovery. c om
i s ys ide
Automatically monitored by an Observerrprocess: u n Gu
bdata ecenter
n t
Located on a distinct server in a distinct @
to Stud
e n
Can be restarted on failure iby
c m Enterprise t h is Manager
Installed through Oracle n s
a Client s e
Administrator
u
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Fast-Startn E s
Failovernis
faefeature that automatically, quickly, and reliably fails over to a
e o tra
rt nsynchronized
designated,
v o -
Ewithoutnrequiring
standby database in the event of loss of the primary database,
manual intervention to execute the failover. In addition, following a fast-start
failover, the original primary database is automatically reconfigured as a new standby
database upon reconnection to the configuration. This enables Data Guard to restore disaster
protection in the configuration as soon as possible.
Fast-Start Failover is used in a Data Guard configuration under the control of the Data Guard
Broker, and may be managed using either dgmgrl or Oracle Enterprise Manager Grid
Control. There are three essential participants in a Fast-Start Failover configuration:
The primary database, which can be a RAC database
A target standby database, which becomes the new primary database following a fast-
start failover
The Fast-Start Failover Observer, which is a separate process incorporated into the
dgmgrl client that continuously monitors the primary database and the target standby
database for possible failure conditions. The underlying rule is that out of these three
participants, whichever two can communicate with each other will determine the
outcome of the fast-start failover. In addition, a fast-start failover can occur only if there
is a guarantee that no data will be lost.

For disaster recovery requirements, install the Observer in a location separate from the
primary and standby data centers. If the designated Observer fails, Enterprise Manager can
detect the failure and can be configured to automatically restart the Observer on the same
host.
You can install the Observer by installing the Oracle Client Administrator (choose the
Administrator option from the Oracle Universal Installer). Installing the Oracle Client
Administrator results in a small footprint because an Oracle instance is not included on the
Observer system. If Enterprise Manager is used, also install the Enterprise Manager Agent on
the Observer system.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

Data Guard Broker Configuration Files
*.DG_BROKER_CONFIG_FILE1=+DG1/orcl/dr1config.dat
*.DG_BROKER_CONFIG_FILE2=+DG1/orcl/dr2config.dat
orcl1 orcl2
s a
)ha
m
co
s
Shared storage u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eof thenData
s feGuard Broker (DGB) configuration files are maintained for each
e o
Two copies
rt son-astrato always have a record of the last-known valid state of the configuration.
v
database
nobroker is started for the first time, the configuration files are automatically created
EWhen the
and named using a default path name and file name that is operating systemspecific.
When using a RAC environment, the DGB configuration files must be shared by all instances
of the same database. You can override the default path name and file name by setting the
following initialization parameters for that database: DG_BROKER_CONFIG_FILE1,
DG_BROKER_CONFIG_FILE2.
You have two possible options to share those files:
Cluster file system
ASM
The example in the slide illustrates a case where those files are stored in an ASM disk group
called DG1. It is assumed that you have already created a directory called orcl in DG1.

Real-Time Query Physical Standby Database
Client Client Client Client

Read/Write Read only
Primary Standby
Redo apply
cluster cluster
s a
Redo apply
instance
)ha
m
co
s
RAC RAC
u n isy uide
database database
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n ERedonApply
s fe (physical standby database) has proven to be a popular solution for
e o
Data Guard
tradue to its relative simplicity, high performance, and superior level of data
rt recovery
v
disaster n -
no Beginning with Oracle Database 11g, a physical standby database can be open
Eprotection.
read-only while redo apply is active. This means that you can run queries and reports against
an up-to-date physical standby database without compromising data protection or extending
recovery time in the event a failover is required. This makes every physical standby database
able to support productive uses even while in standby role. To enable real-time query, open
the database in read-only mode and then issue the ALTER DATABASE RECOVER MANAGED
STANDBY statement. Real-time query provides the ultimate high availability solution because it:
Is totally transparent to applications
Supports Oracle RAC on the primary and standby databases: Although Redo Apply can
be running on only one Oracle RAC instance, you can have all of the instances running
in read-only mode while Redo Apply is running on one instance.
Returns transactionally consistent results that are very close to being up-to-date with the
primary database.
Enables you to use fast-start failover to allow for automatic fast failover in the case the
primary database fails

Hardware Assisted Resilient Data
Blocks validated and

Oracle database
protection information added to blocks

[DB_BLOCK_CHECKSUM=TRUE] Vol Man/ASM
Operating system
Prevents corruption introduced Device driver
in I/O path Host bus adapter
Is supported by major storage vendors:
s a
EMC, Fujitsu, Hitachi, HP, and NEC SAN )h a
Network Appliance &co m
Sun Microsystems i s y s
Virtualization
i d e
r u n Gu
All file types and block sizes checked @ b ent
to StuSAN d
Protection information validated en s interface
c i m h i
n as use t
by storage device when enabled Storage device
symchksum typerto n
Oracle o
tenable
e
v ens e
( e
o n n e2012,
Copyright
v e rt rab
n Ethat ncans e
fcause
r t o
One problem
t r a lengthy outages is data corruption. Today, the primary means for
ve ocorruptions
Edetecting n- caused by hardware or software outside of the Oracle database, such
n subsystem,
as an I/O is the Oracle database checksum. However, after a block is passed to
the operating system, through the volume manager and out to disk, the Oracle database itself
cannot validate whether the block being written is still correct.
With disk technologies expanding in complexity, and with configurations such as Storage
Area Networks (SANs) becoming more popular, the number of layers between the host
processor and the physical spindle continues to increase. With more layers, the chance of any
problem increases. With the HARD initiative, it is possible to enable the verification of
database block checksum information by the storage device. Verifying that the block is still the
same at the end of the write as it was in the beginning gives you an additional level of
security.
By default, the Oracle database automatically adds checksum information to its blocks. These
checksums can be verified by the storage device if you enable this possibility. if a block is
found to be corrupted by the storage device, the device logs an I/O corruption, or it cancels
the I/O and reports the error back to the instance.
Note: The way you enable the checksum validation at the storage device side is vendor-
specific. The example given in the slide uses EMC Symmetrix storage.

Database High Availability: Best Practices
Use SPFILE. Create two or Set Multiplex

more control files. CONTROL_FILE_RECO production and
RD_KEEP_TIME long standby redo
enough. logs
Log checkpoints to Use auto-tune Enable ARCHIVELOG Enable
the alert log. checkpointing. mode and use a flash Flashback
recovery area. Database.
Enable block Use Automatic Use locally managed Use Automatic s a

checking. Undo tablespaces. Segment Space )h a
Management. Management. co m
s
u n
y
isUse u ide
Use resumable Use Database Register all instances
b t G
r ntablespaces
temporary
space allocation. Resource @ tude tempfiles. with
with remote listeners.
t o
Manager.
i m en is S
a s c e th
o n n o us
e r t e t
v
(e 2012, s
eOracle
n n
Copyright l i c
e r t o
a b le
n
v
Ethe s f er
r t o
The table in
t r a n
slide gives you a short summary of the recommended practices that apply to
ve on- databases, RAC databases, and Data Guard standby databases.
Esingle-instance
These n practices affect the performance, availability, and mean time to recover (MTTR) of your
system. Some of these practices may reduce performance, but they are necessary to reduce
or avoid outages. The minimal performance impact is outweighed by the reduced risk of
corruption or the performance improvement for recovery.
Note: For more information about how to set up the features listed in the slide, refer to the
following documents:
Administrators Guide
Data Guard Concepts and Administration
Net Services Administrators Guide

How Many ASM Disk Groups Per Database?
Two disk groups are recommended.

Leverage maximum of LUNs.

Data DG FRA DG
Backups can be stored on one
ERP DB
FRA disk group.
CRM DB
Lower performance may be
HR DB
used for FRA (or inner tracks).
Exceptions: a
a s
Additional disk groups for different capacity or performance
m )h
characteristics
s co
Different ILM storage tiers
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Etime, onlys e
ftwo
Most of
r t othe
t r a n disk groups are enough to share the storage between multiple
ve onThat
Edatabases. - way you can maximize the number of Logical Unit Numbers (LUNs) used as
n which gives you the best performance, especially if these LUNs are carved on the
ASM disks,
outer edge of your disks.
Using a second disk group allows you to have a backup of your data by using it as your
common fast recovery area (FRA). You can put the corresponding LUNs on the inner edge of
your disks because less performance is necessary.
The two noticeable exceptions to this rule are whenever you are using disks with different
capacity or performance characteristics, or when you want to archive your data on lower-end
disks for Information Lifecycle Management (ILM) purposes.

Which RAID Configuration for High Availability?
A. ASM mirroring
B. Hardware RAID 1 (mirroring)
C. Hardware RAID 5 (parity protection)
D. Both ASM mirroring and hardware RAID
Depends on business requirement and budget (cost,has

a
Answer: )
availability, performance, and utilization) om c e
y s
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsyou fe have multiple choices as shown in the slide.
To favor
e t o
rcouldavailability,
a
truse
v
You n -
no Disks)ASM
EInexpensive just mirroring capabilities, or hardware RAID 1 (Redundant Array of
which is a hardware mirroring technique, or hardware RAID 5. The last
possible answer, which is definitely not recommended, is to use both ASM mirroring and
hardware mirroring. Oracle recommends the use of external redundancy disk groups when
using hardware mirroring techniques to avoid an unnecessary overhead.
Therefore, between A, B, and C, the choice depends on your business requirements and
budget.
RAID 1 has the best performance but requires twice the storage capacity. RAID 5 is a much
more economical solution but with a performance penalty essentially for write-intensive
workloads.

Should You Use ASM Mirroring Protection?
Best choice for low-cost storage

Enables extended clustering solutions

No hardware mirroring
s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nthe s festorage array hardware RAID-1 mirroring protection when possible to
e o
Basically, leverage
tra overhead from the server. Use ASM mirroring in the absence of a
rt thenmirroring
v
offload -
noRAID capability.
Ehardware
However, hardware RAID 1 in most Advanced Technology Attachment (ATA) storage
technologies is inefficient and degrades the performance of the array even more. Using ASM
redundancy has proven to deliver much better performance in ATA arrays.
Because the storage cost can grow very rapidly whenever you want to achieve extended
clustering solutions, ASM mirroring should be used as an alternative to hardware mirroring for
low-cost storage solutions.

What Type of Striping Works Best?
A. Only ASM striping (no RAID 0)

B. RAID 0 and ASM striping
C. Use LVM
D. No striping
s a
Answer: A and B ) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
ASM and RAID stripingmare e complementary.
i s S
c i h
n as use t
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Ethe slide,
n s feyou can use only ASM striping, or you can use ASM striping in
e o
As shown in
rt n-withtraRAID 0.
v
combination
no 0, multiple disks are configured together as a set, or a bank, and data from any
EWith RAID
one data file is spread, or striped, across all the disks in the bank.
Combining both ASM striping and RAID striping is called stripe-on-stripe. This combination
offers good performance too.
However, there is no longer a need to use a Logical Volume Manager (LVM) for your
database files, nor is it recommended to not use any striping at all.

ASM Striping Only
Pros: Cons:
Drives evenly distributed for Data & FRA Not well balanced across
Higher bandwidth ALL disks
Allows small incremental growth (73 GB) LUN size limited to disk size
No drive contention
Oracle DB size: 1 TB
Data DG FRA DG
Storage configuration:
1 TB 1673 GB 8arrays with 2 TB 3273 GB s a
LUNs 1273 GB disks per array LUNs
)ha
m
co
s
u n isy uide
b r nt G
o@ tude
RAID 1
n t
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E n s fe slide, you want to store a one-terabyte database with a
e o
In thetcase shown
a in
r n-trtwo-terabytethis
v
corresponding flash recovery area. You use RAID 1 to mirror each disk. In total,
Eyou havenoeight arrays of twelve disks, with each disk being 73 GB. ASM mirroring and
hardware RAID 0 are not used.
In addition, each ASM disk is represented by one entire LUN of 73 GB. This means that the
Data disk group (DG) is allocated 16 LUNs of 73 GB each.
On the other side, the Fast Recovery Area disk group is assigned 32 LUNs of 73 GB each.
This configuration enables you to evenly distribute disks for your data and backups, achieving
good performance and allowing you to manage your storage in small incremental chunks.
However, using a restricted number of disks in your pool does not balance your data well
across all your disks. In addition, you have many LUNs to manage at the storage level.

Hardware RAIDStriped LUNs
Pros: Cons:
Fastest region for Data DG Large incremental growth

Balanced data distribution Data & FRA contention
Fewer LUNs to manage while max
spindles
Data DG FRA DG
1 TB 4250 GB 8arrays with 2 TB 4500 GB s a
)ha
m
co
s
u n isy uide
b r nt G
RAID 0+1
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
e o
In thetcase showna in
r n-trtwo-terabytethis
v
corresponding flash recovery area. You use RAID 0+1, which is a combination of
Ehardwarenostriping and mirroring to mirror and stripe each disk. In total, you have eight arrays
of twelve disks, with each disk being 73 GB. ASM mirroring is not used.
Here, you can define bigger LUNs not restricted to the size of one of your disks. This allows
you to put the Data LUNs on the fastest region of your disks, and the backup LUNs on slower
parts. By doing this, you achieve a better data distribution across all your disks, and you end
up managing a significantly smaller number of LUNs.
However, you must manipulate your storage in much larger chunks than in the previous
configuration.
Note: The hardware stripe size you choose is also very important because you want 1 MB
alignment as much as possible to keep in sync with ASM AUs. Therefore, selecting power-of-
two stripe sizes (128 KB or 256 KB) is better than selecting odd numbers. Storage vendors
typically do not offer many flexible choices depending on their storage array RAID technology
and can create unnecessary I/O bottlenecks if not carefully considered.

Hardware RAIDStriped LUNs HA
Pros: Cons:
Fastest region for Data DG Large incremental growth

Balanced data distribution Might waste space
Fewer LUNs to manage
More high available
Data DG FRA DG
1 TB 2500 GB 8arrays with 1.6 TB 2800 GB s a
)ha
m
co
s
u n isy uide
b r nt G
RAID 0+1
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
e o
In thetcase shown
r n-tr1.6-TB in this
a fast recovery area. You use RAID 0+1, which is a combination of
v
corresponding
Ehardwarenostriping and mirroring to mirror and stripe each disk. In total, you have eight arrays
of twelve disks, with each disk being 73 GB. ASM mirroring is not used.
Compared to the previous slide, you use bigger LUNs for both the Data disk group and the
Fast Recovery Area disk group. However, the presented solution is more highly available than
the previous architecture, because you separate the data from the backups into different
arrays and controllers to reduce the risk of down time if one array fails.
By doing this, you still have a good distribution of data across your disks, although not as
much as in the previous configuration. You still end up managing a significantly smaller
number of LUNs than in the first case.
However, you might end up losing more space than in the previous configuration. Here, you
are using the same size and number of arrays to be consistent with the previous example.

Disk I/O Design Summary
Use external RAID protection when possible.

Create LUNs by using:

Outside half of disk drives for highest performance
Small disk, high rpm (that is, 73 GB/15k rpm)
Use LUNs with the same performance characteristics.
Use LUNs with the same capacity.
Maximize the number of spindles in your disk group. s a
)ha
m
co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
Use ASM n E nsand
for volume
fe file management to equalize the workload across disks and
e o
rt hot tra The following are simple guidelines and best practices when configuring
eliminate
v n -spots.
nogroups:
EASM disk
Use external RAID protection when possible.
Create LUNs using:
- The outside half of disk drives for highest performance
- Small disks with high rpm (for example, 73 GB with 15k rpm). Speed is important,
because it impacts both positioning time and data transfer. This means that faster
spindle speed drives have improved performance regardless of whether they are
used for many small, random accesses, or for streaming large contiguous blocks
from the disk. The stack of platters in a disk rotates at a constant speed. The drive
head, while positioned close to the center of the disk, reads from a surface that is
passing by more slowly than the surface at the outer edges.
Maximize the number of spindles in your disk group.
LUNs provisioned to ASM disk groups should have the same storage performance and
availability characteristics. Configuring mixed speed drives will default to the lowest
common denominator.
ASM data distribution policy is capacity based. Therefore, LUNs provided to ASM
should have the same capacity for each disk group to avoid imbalance and hot spots.

Extended RAC: Overview
Full utilization of resources, no matter where they are

located
Site A RAC Site B

database
Clients s a
)ha
m
co
s
Site A RAC
u n isy uSiteideB
database
b r nt G
n t o@ tude
Fast recovery from site failure
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E databases
n s fe share a single set of storage and are located on servers in the
e o
Typically, RAC
tra
rtdatancenter.
v
same -
no RAC, you can use disk mirroring and Dense Wavelength Division Multiplexing
EWith extended
(DWDM) equipment to extend the reach of the cluster. This configuration allows two data
centers, separated by up to 100 kilometers, to share the same RAC database with multiple
RAC instances spread across the two sites.
As shown in the slide, this RAC topology is very interesting, because the clients work gets
distributed automatically across all nodes independently of their location, and if one site goes
down, the clients work continues to be executed on the remaining site. The types of failures
that extended RAC can cover are mainly failures of an entire data center due to a limited
geographic disaster. Fire, flooding, and site power failure are just a few examples of limited
geographic disasters that can result in the failure of an entire data center.
Note: Extended RAC does not use special software other than the normal RAC installation.

Extended RAC Connectivity
Distances over ten kilometers require dark fiber.

Set up buffer credits for large distances.
Dark fiber
Site A Site B
DWDM DWDM
device device
s a
)ha
co m
s
DB
u
DB isy
n u ide
copy
b rcopy nt G
n t o@ tude
mePublic i s S
network
c i h
Clients
n as use t
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E na sRAC fe cluster to another site separated from your data center by more than
e o
In order to extend
rt n-tritais required to use DWDM over dark fiber to get good performance results.
v
ten kilometers,
EDWDMnisoa technology that uses multiple lasers, and transmits several wavelengths of light
simultaneously over a single optical fiber. DWDM enables the existing infrastructure of a
single fiber cable to be dramatically increased. DWDM systems can support more than 150
wavelengths, each carrying up to 10 Gbps. Such systems provide more than a terabit per
second of data transmission on one optical strand that is thinner than a human hair.
As shown in the slide, each site should have its own DWDM device connected together by a
dark fiber optical strand. All traffic between the two sites is sent through the DWDM and
carried on dark fiber. This includes mirrored disk writes, network and heartbeat traffic, and
memory-to-memory data passage. Also shown in the graphic are the sets of disks at each
site. Each site maintains a copy of the RAC database.
It is important to note that depending on the sites distance, you should tune and determine
the minimum value of buffer credits in order to maintain the maximum link bandwidth. Buffer
credit is a mechanism defined by the Fiber Channel standard that establishes the maximum
amount of data that can be sent at any one time.
Note: Dark fiber is a single fiber optic cable or strand mainly sold by telecom providers.

Extended RAC Disk Mirroring
Need copy of data at each location

Two options:
Host-based mirroring
Remote array-based mirroring
Recommended solution with ASM
Site A Site B Primary Secondary

s a
)ha
m
co
s
u n isy uide
b r nt G
t o @ tude
n DB S
DB
copy
DB
copy c i me tcopy h is DB
copy
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E is only
n s fe
Although
r t o there
t r a one RAC database, each data center has its own set of storage that is
ve on- mirrored using either a cluster-aware, host-based Logical Volume Manager
Esynchronously
n
(LVM) solution, such as SLVM with MirrorDiskUX, or an array-based mirroring solution, such
as EMC SRDF.
With host-based mirroring, shown on the left of the slide, the disks appear as one set, and all
I/Os get sent to both sets of disks. This solution requires closely integrated clusterware and
LVM, and ASM is the recommended solution.
With array-based mirroring, shown on the right, all I/Os are sent to one site, and are then
mirrored to the other. In fact, this solution is like a primary/secondary site setup. If the primary
site fails, all access to primary disks is lost. An outage may be incurred before you can switch
to the secondary site.
Note: With extended RAC, designing the cluster in a manner that ensures the cluster can
achieve quorum after a site failure is a critical issue. For more information regarding this topic,
refer to the Oracle Technology Network site.

Achieving Quorum with Extended RAC
Third
site
Voting disk (NFS or iSCSI)
Redundant public network
Redundant private network

s a
)Bha
Site A
m
co
Site
s
Redundant SAN with ASM mirroring
u n isy uide
RAC b r RAC n t G
database t o@ tuddatabase e
n
Database files (ASM failure group) c i me thDatabase is S files (ASM failure group)
OCR
n as use OCR
Voting disk
r t o n t o Voting disk
( e ve ense
o n n e2012,
Copyright
v e rt rab
n E disks e
sfare
r o
As fartas voting a
t r n concerned, a node must be able to access strictly more than half of
vevotingodisks
the - at any time, or that node will be evicted from the cluster. Extended clusters
Eare n n implemented with only two storage systems, one at each site. This means that
generally
the site that houses the majority of the voting disks is a potential single point of failure for the
entire cluster. To prevent this potential outage, Oracle Clusterware supports a third voting disk
on an inexpensive, low-end, standard NFS-mounted device somewhere on the network. It is
thus recommended to put this third NFS voting disk on a dedicated server visible from both
sites. This situation is illustrated in the slide. The goal is that each site can run independently
of the other when a site failure occurs.
Note: For more information about NFS configuration of the third voting disk, refer to the
Oracle Technology Network site.

Additional Data Guard Benefits
Greater disaster protection

Greater distance
Additional protection against corruptions
Better for planned maintenance
Full rolling upgrades
More performance neutral at large distances
Option to do asynchronous transfer s a
) ha
If you cannot handle the costs of a DWDM network,oData m
Guard still works over cheap, standard networks. y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n Eprovidesn s faegreater disaster protection:
e t o
Data Guard
rDistance a
trover
E v o n - 100 kilometers without a performance hit
n
Additional protection against corruptions, because it uses a separate database
Optional delay to protect against user errors
Data Guard also provides better planned maintenance capabilities by supporting full rolling
upgrades.
In addition, if your budget cannot handle the costs of a DWDM network, you can still use Data
Guard because it works over inexpensive, standard networks.

Using a Test Environment
The most common cause of down time is change.

Test your changes on a separate test cluster before

changing your production environment.
Production Test s a
cluster cluster )ha
m
co
s
u n isy uide
b r nt G
RAC
n t o@RACtude
database
c i me thdatabase is S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n fe cause of down time in a production environment. A proper test
E mostnslikely
Change
e o is the
rt n-can tracatch more than 90 percent of the changes that could lead to a down time of
v
environment
no environment, and is invaluable for quick test and resolution of issues in
Ethe production
production.
When your production environment is RAC, your test environment should be a separate RAC
cluster with all the identical software components and versions.
Without a test cluster, your production environment will not be highly available.
Note: Not using a test environment is one of the most common errors seen by Oracle Support
Services.

Quiz
Which of the following statements regarding Disk I/O design are

true?
a. Use external RAID protection when possible.
b. Use LUNs with the same performance characteristics.
c. Use LUNs with the same capacity.
d. Minimize the number of spindles in your disk group.
s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E c nsfe
Answer:
e o a, b,
rt n-tra
v
E no

Summary

Design a Maximum Availability Architecture in your

environment
Determine the best RAC and Data Guard topologies for
your environment
Configure the Data Guard Broker configuration files in a
RAC environment s a
ha
Identify successful disk I/O strategies m) co
s
u n isy uide
b r nt G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e2012,
Copyright
v e rt rab
n E nsfe
e o
rt n-tra
v
E no

s a
) ha
o m
y s c e
u n is uid
b r n t G
n t o@ tude
c i me this S
n as use
e r ton e to
( e v ens
o n n e lic
e r t a b l
n Ev nsfer
e r to -tra
Ev non

d50311gc40 Rac Admin SG

Uploaded by

Copyright:

Available Formats

d50311gc40 Rac Admin SG

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

d50311gc40 Rac Admin SG

Uploaded by

Copyright:

Available Formats

Unauthorized reproduction or distribution prohibited Copyright 2013, Oracle and/or its affiliates

James Womack Disclaimer

1 Grid Infrastructure: Overview

Performance of Typical Components 2-17

3 Installing and Configuring Oracle RAC a

4 Oracle RAC Administration

Enterprise Manager Metrics and RAC 4-11

5 Managing Backup and Recovery for RAC

Channel Connections to Cluster Instances 5-21

Typical Latencies for RAC Operations 7-6

8 Managing High Availability of Services

Using Services with Client Applications 8-11

Load Balancing Advisory 9-22

Performing an Online Migration 11-6

Oracles Solution to Down Time 13-5

Grid Infrastructure: Overview

After completing this lesson, you should be able to:

Explain the principles and purposes of clusters

Oracle Database 11g: RAC Administration 1 - 2

Oracle Database 11g: RAC Administration 1 - 3

A group of independent, but interconnected, computers

that act as a single system

Oracle Database 11g: RAC Administration 1 - 4

Software that provides various interfaces and services for a

cluster. Typically, this includes capabilities that:

Oracle Database 11g: RAC Administration 1 - 5

Oracle Clusterware is:

A key part of Oracle Grid Infrastructure

Oracle Database 11g: RAC Administration 1 - 6

Shared disk cluster architecture supporting application

load balancing and failover

Oracle Database 11g: RAC Administration 1 - 7

Oracle Database 11g: RAC Administration 1 - 8

The interconnect adapter must support:

Oracle Database 11g: RAC Administration 1 - 9

Oracle Database 11g: RAC Administration 1 - 10

Platform-independent facility for starting, stopping, and

Oracle Database 11g: RAC Administration 1 - 11

Oracle Database 11g: RAC Administration 1 - 12

cssdagent orarootagent oraagent cssdmonitor s a

Oracle Database 11g: RAC Administration 1 - 13

Oracle Database 11g: RAC Administration 1 - 14

cssdagent orarootagent oraagent cssdmonitor

Oracle Database 11g: RAC Administration 1 - 15

Oracle Database 11g: RAC Administration 1 - 16

cssdagent orarootagent oraagent cssdmonitor

cssdagent orarootagent oraagent cssdmonitor

Oracle Database 11g: RAC Administration 1 - 18

orarootagent cssdagent oraagent cssdmonitor

Oracle Database 11g: RAC Administration 1 - 19

In previous releases, adding or removing servers in a

cluster required extensive manual preparation.

Oracle Database 11g: RAC Administration 1 - 20

GNS is an integral component of Grid Plug and Play.

The only static IP address required for the cluster is the

Oracle Database 11g: RAC Administration 1 - 21

The SCAN is a fully qualified host name located in the GNS

Oracle Database 11g: RAC Administration 1 - 22

scan.GNS_domain. If a GNS domain is not specified at installation, the SCAN defaults to