B-Ha For Mca Vcs Hadr WP 14187067.en-Us

W
H
I
T
E

P
A
P
E
R
:

C
U
S
T
O
M
I
Z
E
Confidence in a connected world.
W
H
I
T
E

P
A
P
E
R
:

T
E
C
H
N
I
C
A
L

High Availability for Mission-
Critical Applications
Protecting Applications with
Veritas Cluster Server
Eric Hennessey, Director
Technical Product Management
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
The challenge of application availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Veritas Cluster Server from Symantec . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Key features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Components of Veritas Cluster Server with application support . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
How Veritas Cluster Server manages an application. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Anatomy of an application failover with Veritas Cluster Server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
How Veritas Cluster Server addresses customer challenges. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
The need to control costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
N+1 roaming spare clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
N to N clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
The need to reduce complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
The need to manage dependent, multi-tier applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
The need to address disaster recovery requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Managing multiple clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Additional benefits of Veritas Storage Foundation HA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
White Paper: Technical
High Availability for Mission-
Critical Applications
Protecting Applications with
Veritas Cluster Server
High Availability for Mission-Critical Applications:
Protecting Applications with Veritas Cluster Server
4
Introduction
In years past, IT solutions were deployed to support back-end business functions such as
inventory management, accounting, and human resources. In todays world, the IT solution
increasingly is the business. The importance of availability for services from back-end database
servers, through mid-tier application servers to front-end web servers has never been greater.
This paper will discuss the various options available for providing high availability (HA)
for applications offered by Veritas Cluster Server and describe how Veritas Cluster Server from
Symantec keeps applications highly available.
The challenge of application availability
Organizations that manage mission-critical applications are concerned with keeping those
applications as available as possible. However, the concern is complicated by the need to:
Control costs
Reduce complexity
Address dependent, multi-tier applications
Address disaster recovery requirements
Traditional high-availability failover solutions do not provide parallel access, but can provide
a very short recovery time objective (RTO) at significantly reduced cost and complexity.
Operating system vendors and several third-party software vendors provide classic high-
availability failover products. Most of these products were designed for, or operate best in,
a traditional active-passive failover configuration. In this configuration, the active node or
server runs an application, and the passive failover node stands idle, waiting for failure. This
configuration does help sell additional servers, but it is not a cost-effective solution from the
customer standpoint.
Because most companies do not operate a single mission-critical application but, rather,
a large number of them, there is a cost-effective solution that provides solid, dependable high
availability without requiring the expense of provisioning a spare server for every mission-critical
application. The Veritas Cluster Server high-availability solution from Symantec is designed to
provide cost-effective roaming spare or no spare failover configurations for mission-critical
applications.
5
Veritas Cluster Server from Symantec
Veritas Cluster Server from Symantec is the industry-leading cross-platform high-availability
clustering solution. With its application-centric approach to availability, Veritas Cluster Server
provides customers a high degree of flexibility in determining the application failover behavior
best suited to their business requirements.
Veritas Cluster Server provides high availability to mission-critical applications by managing
each application within a construct known as a service group. Each service group contains the
specific application components needed to manage an application, along with all of the network
and storage resources required by it. The service group is the single unit of failover within a
Veritas Cluster Server cluster.
Key features
Veritas Cluster Server is a feature-rich, best-in-class high availability clustering solution. Key
features include:
Intelligent failover logic with service group workload management
Out-of-the-box support for most major applications
Automated disaster recovery (DR)
Disaster recovery and local high-availability failover test suite
Management console for multi-cluster and multi-site management
Components of Veritas Cluster Server with application support
Clusters
A single Veritas Cluster Server cluster consists of multiple systems connected in various
combinations to shared storage devices. Veritas Cluster Server monitors and controls applications
running in the cluster, and can restart applications in response to a variety of hardware or
software faults.
A cluster is defined as all systems that share a common cluster configuration and utilize
a common interconnect network. The Veritas Cluster Server cluster interconnect consists
of redundant physical Ethernet interconnects, generally over two or more dedicated private
networks. The communications layer carries heartbeats between systems within the cluster, as
well as membership and state change information.
6
Clusters can have from one to 32 member systems, or nodes. Within a single Veritas Cluster
Server cluster, all member nodes must run the same operating system family. For example, a
Solaris cluster would consist of entirely Solaris nodeslikewise with Linux, AIX, HP-UX,
and Windows clusters. Within a given cluster, multiple versions of the operating system can be
deployed to support specific upgrade scenarios.
Applications can be configured to run on specific nodes in the cluster. Storage is configured
to provide access to shared application data for the systems that are hosting the application. In
that respect, the actual storage connectivity will determine where applications can be run: Nodes
sharing access to storage are eligible to run an application.
Agents
Veritas Cluster Server agents handle the start, stop, and monitoring of all resources contained
within a service group. Agents receive instructions on when to start, stop, or monitor a resource
from the Veritas Cluster Server engine, and the agents then return the results of those actions to
the engine. Bundled with the Veritas Cluster Server package are a collection of agents to manage
the storage and network resources required by an application managed by Veritas Cluster Server.
Veritas Cluster Server also ships with agents to control all common system functions, such
as file systems and network addresses. Additional agents are provided for out-of-the-box support
for most enterprise applications, such as databases, application servers, and Web servers. This
includes complete out-of-the-box (no customization required) support for Oracle, DB2, Sybase,
WebSphere, WebLogic, and many other enterprise applications.
How Veritas Cluster Server manages an application
Veritas Cluster Server support for a given application consists of an agentor, in the case of more
complex applications such as SAP, a collection of agents. The agent responsible for the application
(or application component) takes care of starting, stopping, and monitoring the health of the
application.
To control the application, the agent starts and stops the application by calling the
applications defined start or stop procedure, which can be a script or executable binary.
Monitoring of the application can be performed in a number of ways, ranging from checking the
process table for one or more application processes to connecting to the application and testing
its ability to communicate.
:
7
Anatomy of an application failover with Veritas Cluster Server
To understand the sequence of events that occurs when Veritas Cluster Server fails-over an
application, it is helpful to understand how a Veritas Cluster Server service group is configured
for the application. Figure 1 below is from the Veritas Cluster Server console, and it illustrates a
typical Oracle service group configuration in Veritas Cluster Server:
Figure 1: Oracle service group, as viewed from the Veritas Cluster Server console.
Each blue box in the figure represents a resource managed by Veritas Cluster Server within
this service group. At the top is the Oracle Listener service, and beneath that on the right is
the Oracle database instance itself. The remaining boxes represent the storage and network
resources; the virtual IP and NIC resources on the left, and the file systems and Veritas Volume
Manager disk group on the right. The virtual IP address managed by Veritas Cluster Server here is
the IP address to which the application is bound. Each of these resources is managed by a specific
agent provided with Veritas Cluster Server.
8
The lines connecting the resources show the various dependencies between the resources,
which dictate the correct start and stop sequence for all the resources in the service group. For
example, the Oracle database should not be started before the file systemswhich contain its
archives and database filesare mounted.
In a planned switch-over eventi.e., one in which a conscious decision has been made to
move the Oracle instance from one system to another in the clusterthe following sequence of
events will take place once the command is issued:
1. The Oracle listener agent stops the listener service.
2. The virtual IP address to which the listener was bound is removed from its network
interface and, at the same time, the Oracle instance is shut down.
3. All three file systems (archive, DB files, and redo logs) are unmounted.
4. The disk group containing the file systems is deported.
For the target nodethe node chosen by the administrator to host the failover databasethe
above sequence is reversed.
Of course, not all failover events are planned. Depending on the nature of the problem that
caused the failover, Veritas Cluster Server may or may not be able to gracefully shut down all the
resources within the service group. In the case of a server panic or power outage, none of the
resources will be brought down cleanly and, when startup occurs on a new node, standard crash
recovery will take place. When the service group is started on the new node, the file system agent
will take care of performing necessary file system checks before mounting, and the application
will perform its standard recovery on startup. It is vital to understand that application recovery
on a failover is identical to normal application startup following any local server issue. In order
to recover quickly, it is critical that you utilize a journaled file system, such as the Veritas File
System, and that your application be capable of restarting cleanly and quickly following a crash.
A database such as Oracle is an example of an application that recovers to a known state quickly
following a crash.
9
How Veritas Cluster Server addresses customer challenges
Beyond the simple ability to move an application from one system to another in response to a
failure, Veritas Cluster Server also provides significant additional benefits to the customer. These
include addressing the problem areas of controlling costs, reducing complexity, and providing
high availability to dependent (multi-tier) applications.
In this section, we will explore the question of how Veritas Cluster Server addresses these
customer challenges and how it provides automated disaster recovery.
The need to control costs
Traditional approaches to high availability have entailed the clustering of mission-critical
applications with pairs of servers in active-passive (standby) configurationwhich works well
enough in a smaller environment with only one or two mission-critical applications. But in a
larger environment with many such applications, the cost of the required hardware spares and
additional software licenses is prohibitive. Veritas Cluster Server allows the deployment of large
clusters with advanced failover logic that can make intelligent failover target selections based
on server capacity and application load. With Veritas Cluster Server, all of a companys servers
that are running mission-critical applications can be grouped together in a single cluster. Then,
one or two spare systems can easily be added to that cluster to handle failovers. This is called
N+1 clustering, where N represents the number of servers with active applications and 1 indicates
a single spare. In a larger cluster, this might actually be N+2 or N+3.
A variant of this cluster topology is N to N. In this topology, there is no spare server per
se, but each server has some amount of spare capacity to host additional applications. The
paragraphs that follow will discuss these topologies in more detail.
N+1 roaming spare clusters
The N+1 cluster topology has a number of active servers, plus one idle server as a spare. For this
example, we will consider a three-node cluster with two application instances, Application A and
Application B, which are active on two separate servers, Server 1 and Server 2. Server 3 is, initially
at least, the spare server.
At the initial cluster startup, Application A is started on Server 1 and Application B is started
on Server 2. Server 3 is idle. If a fault occurs on Server 1, Application A fails-over to Server 3.
When Server 1 is returned to service, there is no automatic fail-back of Application A, as a fail-
back would introduce an additional unnecessary outage. Instead, Server 1 simply becomes the
new spare. This is the meaning of the term roaming spare. Veritas Cluster Server will always
10
choose the server that is currently idle for failover, without requiring an update to the failover
order in the cluster configuration.
In larger N+1 clusters, the intelligent failover logic that Veritas Cluster Server offers with its
service group workload management (SGWM) capability becomes evident. If you start with an N+1
configuration and you lose the floating spare server, Veritas Cluster Server is intelligent enough
to properly handle making the best possible choice from the remaining running nodes. This
management capability associates a numerical load value with each service group (application)
and a numerical capacity value with each server. The cluster considers and evaluates these
values when it selects a failover target. In the absence of an intelligent method of selecting a
target server for failoverssuch as SGWMin the event of multiple failures, applications tend to
become stacked onto one or two servers.
N to N clusters
The N to N cluster topology does not include a spare server. All systems that host a mission-
critical application are placed into the cluster with the assumption that each server has some
extra capacity to host additional applications in the event of a failure.
As with N+1 clusters, each service group (application) is assigned a numerical load value
and each server in the cluster a numerical capacity value. Because the cluster has no idle servers,
the cluster must be able to select failover target nodes based on load and capacity; conventional
ordered list selection simply will not work.
This topology also lends itself well to server consolidation efforts, because the cluster will
automatically attempt to stack applications on all of the clusters servers in a balanced fashion.
Veritas Cluster Server provides an additional construct known as limits (for servers) and
prerequisites (for service groups). This permits a user to define the upper limits of a servers
resources and the minimum amount of those resources that must be reserved for the application
it hosts. For example, consider a case where the user has configured adequate shared memory,
semaphores, etc., in the kernel configuration to host a maximum of two running databases on
each of four servers. The user can then define a limit value of Databases=2 for each server.
Each service group hosting a database can then have a prerequisite value of Databases=1.
When a service group that is hosting a database comes online on one of the database servers,
the service groups prerequisite value is deducted from the servers limit value. If no server exists
with an adequate limit value remaining, the service group is not allowed to come online. By using
11
a combination of capacity vs. load and limits vs. prerequisites, the user can create an automated
system to choose the best possible host in a failover, yet prevent any specific server from being
overloaded. This capability is unique to Veritas Cluster Server; it is not possible to do this with
traditional cluster products that use a simple list of servers for failover decision-making.
In contrast to this capability, other cluster products select a failover target from a simple
ordered list of servers. In general, this list is set so that an application runs on one server in
the cluster under normal circumstances, but will move to a designated spare upon failover.
This practice is fairly adequate until subsequent failures cause the designated spare to become
overloaded as multiple applications fail-over to it.
The need to reduce complexity
Veritas Cluster Server reduces complexity in a number of ways. Because of its consistency across
OS platforms, Veritas Cluster Server reduces the administrative burden by providing a consistent
clustering solution across the enterpriseso systems administrators and operations staff have
only a single user interface to learn, regardless of the OS platform in use.
Because Symantec provides a wide selection of pre-built agents to manage most popular
enterprise applications, no custom scripting is needed to get a supported application under
Veritas Cluster Server control. To cluster a given application, the user must simply supply a few
pieces of information about the application in order for Veritas Cluster Server to manage it.
Veritas Cluster Server also ships with a fully functional Cluster Simulator for testing and
training. Customers can use the Cluster Simulator to test new cluster configurations and changes
prior to rolling them out in production. The Cluster Simulator also provides a perfect environment
for new operator and administrator training. The Veritas Cluster Server Cluster Simulator can be
installed on any Windows desktop system and is available on the Symantec corporate Web site.
The need to manage dependent, multi-tier applications
Todays applications are rarely islands unto themselves. Usually, an application is part of a multi-
tier application stack that also encompasses databases and Web tiers. When an application is
started, all applications comprising the multi-tier stack need to be started in a proper sequence.
In most cases, when a back-end database fails-over, it is likely that other dependent applications
either will have to fail-over as well or will have to be restarted in order to re-establish connections.
12
In cases where the various tiers are running on separate OS platforms, Veritas Cluster
Server provides a remote group agent to establish cross-cluster service group dependencies. This
capability is key to multi-tier application control, and it is unique to Veritas Cluster Server.
The remote group agent is configured to point to a service group running on another cluster
in the data center. When a failover occurs, the remote group resource will wait for the service
group in the remote cluster to come online before bringing the local service group online. This
enables the startup of applications in a proper sequence even when the applications are running
on separate OS platforms.
The need to address disaster recovery requirements
In the event of a disaster that causes a complete data center outage, it is imperative that mission-
critical applications be brought online as quickly as possible at an alternate site.
Veritas Cluster Server HA/DR enables global clustering, which extends the concepts of local
high availability to wide-area disaster recovery failover. In a Veritas Cluster Server global cluster
configuration, a Veritas Cluster Server cluster is deployed at each of two to four data centers.
Communications are established between the clusters over an IP network, and each cluster is
configured with an identical service group. Keep in mind that the primary Veritas Cluster Server
cluster can have multiple service groups. In combination with the global clustering feature of
Veritas Cluster Server, this capability provides flexibility in replicating the critical applications
of the businesswhich can significantly reduce the cost of the total disaster recovery solution.
The application data is replicated between storage devices at each site, and the service group
is capable of failing-over not only within its own local cluster in response to a localized failure,
but also to another cluster in the event of complete loss of the entire data center. Veritas Cluster
Server HA/DR includes agents for most enterprise storage replication technologies (e.g., EMC
SRDF and IBM GlobalMirror) and Veritas Volume Replicator, as well as DB2 HA/DR. The idea
behind the disaster recovery capabilities of Veritas Cluster Server is to protect the application, not
just the data.
In order to provide regular validation that a wide-area failover will work when the time
comes, Veritas Cluster Server provides a testing feature called Fire Drill that allows for non-
disruptive testing of the disaster recovery configuration. When a test is run, Veritas Cluster
Server creates a storage snapshot of the application data at the secondary data center and then
brings the application up at the secondary data center without its network identity. (The network
13
components are omitted to prevent interference with the production application running at the
primary data center.) The user can use this remote replica of the data to verify that the application
can be started successfully. Because most disaster recovery failures stem from changes to the
primary sites storage configuration that are not propagated to the secondary site, the completion
of a Fire Drill test provides assurance that all required storage is being replicated and is valid.
Once testing is complete, the application is brought back down and the snapshot is released.
The production application running at the primary data center is completely unaffected by the
Fire Drill test at the secondary site.
In a real disaster scenario or planned application movement, the process is only slightly
different. The actual application data replica is mounted, and all related network components
such as the IP address and any DNS updatesare started so that customers can continue to
operate.
In the case of a planned application movement from one data center to another, the direction
of replication is reversed so that data consistency is maintained at both sites.
Managing multiple clusters
In larger enterprises, it is common for multiple clusters to be deployed not just across multiple
OS platforms, but across multiple data centers as well. This practice can pose an administrative
challenge, especially when dependencies exist between applications running in different clusters
or when applications are capable of failing-over across data centers. In response to this challenge,
Symantec offers a consolidated Veritas Cluster Server Management Console, which provides full
visibility into and management of all Veritas Cluster Server clusters throughout the enterprise,
across multiple OS platforms and across multiple sites.
The console permits management and operational control over all Veritas Cluster Server
clusters from a single pane of glass. In addition, it further extends wide-area disaster recovery
management capabilities by providing a simplified site migration function that allows all
applications under Veritas Cluster Server control to be relocated from one data center to another.
A rich set of pre-defined reports, such as application uptime reports for SLA compliance, can
be produced from the Management Consoleand it can also provide complex metropolitan- and
wide-area cluster visualizations.
The Management Console was introduced starting with the Veritas Cluster Server 5.0 release
and is included as part of the base product. It can be downloaded from the Veritas Cluster Server
product page at http://go.symantec.com/vcs.
14
Additional benefits of Veritas Storage Foundation HA
When Veritas Cluster Server is bundled with Veritas Storage Foundation, applications derive
additional benefits in the form of increased performance, flexibility, and reliability.
Veritas Storage Foundation consists of the following components:
Veritas Volume ManagerVeritas Volume Manager provides simplified disk management
with striping and mirroring for improved performance and availability. This includes the
ability to mirror across sites within metropolitan area cluster (MAC) distances of about 50
miles. Dynamic multi-pathing (DMP) provides availability and load balancing across storage
connections (host bus adapters [HBAs]).
Veritas File SystemVeritas File System is a high-performance, fast-recovery file system
that is optimized for business-critical database applications and data-intensive workloads.
Veritas File System offers online administration, letting you perform most frequently
scheduled maintenance tasks (including online backup, resizing, and file system changes)
without interrupting data or system availability. Veritas File System also provides support for
large file systems (of more than eight exabytes in a 64-bit environment) and large files (in the
exabyte range in a 64-bit environment). Veritas File System offers the following performance-
enhancing features that are of particular interest in a database environment:
Veritas Quick I/OImproves the throughput databases built on the Veritas File System. It
delivers raw device performance to databases, providing the administrative advantages of
using file systems without the performance penalties.
Veritas Cached Quick I/OFurther enhances database performance by leveraging large
system memory to selectively buffer frequently accessed data.
Veritas Concurrent I/OImproves the performance of regular file systems without the
need to extend namespaces and present the files as devices. This simplifies administrative
tasks and allows relational databases that have no sequential read/write requirement to
access files concurrently.
Veritas Enterprise AdministratorVeritas Enterprise Administrator is the infrastructure
that allows Veritas Storage Foundation, Veritas Volume Manager, and Veritas File System
information and features to be accessed through the graphical user interface (GUI).
15
Summary
Veritas Cluster Server from Symantec offers the highest degree of flexibility and cost-effectiveness
in protecting mission-critical applications from both localized and data centerwide failures.
When deployed in an enterprise-wide environment, Veritas Cluster Server protects applications,
along with their dependent applications, across multiple OS platforms and multiple locations.
When Veritas Cluster Server is deployed in conjunction with Veritas Storage Foundation,
unparalleled levels of reliability, performance, and manageability are possible.
About Symantec
Symantec is a global leader in
providing security, storage, and
systems management solutions to
help businesses and consumers
secure and manage their information.
Headquartered in Cupertino, Calif.,
Symantec has operations in more than
40 countries. More information is
available at www.symantec.com.
Copyright 2007, 2008 Symantec Corporation. All
rights reserved. Symantec, the Symantec Logo, Veritas,
and Veritas Storage Foundation are trademarks or
registered trademarks of Symantec Corporation or its
affiliates in the U.S. and other countries. HP and HP-UX
are trademarks of Hewlett-Packard Company. IBM,
AIX, and DB2 are trademarks of International Business
Machines Corporation in the United States, other coun-
tries, or both. Linux is the registered trademark of Linus
Torvalds. Microsoft and Windows are either registered
trademarks or trademarks of Microsoft Corporation
in the United States and/or other countries. Sun and
Solaris are trademarks or registered trademarks of Sun
Microsystems, Inc., in the U.S. or other countries. Other
names may be trademarks of their respective owners.
8/08 14187067
For specific country offices and
contact numbers, please visit
our Web site. For product
information in the U.S., call
toll-free 1 (800) 745 6054.
Symantec Corporation
World Headquarters
20330 Stevens Creek Boulevard
Cupertino, CA 95014 USA
+1 (408) 517 8000
1 (800) 721 3934
www.symantec.com

B-Ha For Mca Vcs Hadr WP 14187067.en-Us

Uploaded by

Copyright:

Available Formats

B-Ha For Mca Vcs Hadr WP 14187067.en-Us

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

B-Ha For Mca Vcs Hadr WP 14187067.en-Us

Uploaded by

Copyright:

Available Formats

W

You might also like