Steve Klassen

Here are my thoughts on #observability, in no particular order: If your application doesn't produce any #logs (#metrics, #spans, you name it), troubleshooting becomes impossible. You may not even know if it's running. This is especially problematic when your application logs only during usage, but it's not being used. On the flip side, if your application generates too many logs, it can also be challenging to troubleshoot, even if you have log querying tools. Logging "Health check succeeded!" every second for routine health checks contributes to log overload, compounding the previous issue. When log entries are excessively long (structured or not, someone will have to look at your logs at some point). When structured logs contain large, embedded JSON objects. When structured logs embed lengthy lists with line breaks (which are encoded). When stack traces dive into the internals of your HTTP server (or any other third-party library). When the same information is logged repeatedly across 15 layers of error handling. When you cannot attribute an error to an invocation or input parameters. When you log sensitive information. Happy logging!

8

As a long-time Debian user, I never thought I’d make the switch—until I tried Pop!_OS. While Debian is incredibly solid and secure. I loved the stability and simplicity of Debian, Pop!_OS brings a modern touch with its ease of use and extra features—without sacrificing performance. One key feature is the built-in tiling window manager. I used to install Xmonad on debian to get the tiling manager and it needs a lot of config changes/tweaks. Have you considered switching distros recently? Let’s chat! #Pop_OS #Debian #Linux #TilingWindowManager #Productivity

16

2 Comments

Learn how to use Kyverno, the native policy engine for Kubernetes, to enforce policies, validate configurations, and manage resources dynamically within your Kubernetes clusters. https://lnkd.in/gaDZkFm9

48

2 Comments

Important checklist 📋 Check comments as well for more useful tips. #cloud #aws #security #compliance #cost_management

7

I have been wondering--my provisioned self-hosted GitLab runner on kubernetes is extremely faster than their SaaS runner. In fact by orders of magnitude. I provisioned a Group top-level runner inherited by all sub-group and projects--and that’s enough for me. When I worked at compare the market, we provisioned runner per-function and per squads and that’s because of security considerations and to avoid noisy neighbours runners. Some Java/android team runners were extremely hungry that I had to provision runners with over 32GB memory capacity for and also looking at accelerator runners for them. #kubernetes #cloudengineering #cloudcomputing #gitlb

20

3 Comments

One thing I have been noticing over the years is how people confuse observability with just monitoring. If you really want to get beyond just mom your infrastructure then you really need to listen to what your monitoring is actually telling you. I have set up many an instance of Grafana, people love the fancy charts and graphs but very few people really listen to them. Take it beyond the fancy charts to notice the patterns that you are seeing in them and using that to be proactive before you have to be reactive.

3

2 Comments

The latest addition to https://lnkd.in/gh628sFJ has landed to teach you how #jsonpath RFC from @ietf works! Go check it out! #jsoneverything

2

Hopefully I can save someone time. Microsoft Kubernetes extension for VS Code stopped working for me today, would not show anything in the panels except the message: "There is no data provider registered that can provide view data." After much troubleshooting and gnashing of teeth, I noticed version 1.3.17 was just released. Installing specific version 1.3.16 fixed the problem for me. UPDATE: I upgraded it to 1.3.18 and it is working fine. Perhaps a bug was fixed?

5

1 Comment

Honestly, I originally thought this was "Not Good For Work", but it turns out that it's "Next Generation Firewall". Anyway, I guess it helps you keep out the stuff that is not good for work? Check out this AWESOME 14-minute Cloud NGFW Firewall Enterprise overview and demo to get a hands-on look at how to: - Set up Cloud NGFW Enterprise. - Create rules to block malicious traffic. - Monitor your network for threats. This is a great resource for anyone who wants to learn more about how Cloud NGFW Enterprise, powered by @PaloAltoNetworks, can help you secure your cloud environment. Watch now: https://bit.ly/3XdKGEa #cloudsecurity #googlecloud #PaloAltoNetworks #NGFW

3

OpenTelemetry might be in beta, some parts are newer than others, and the configuration might be daunting. But that’s nothing compared to ripping apart your codebase if your SaaS vendor hikes the prices. Abstracting your vendor isn’t just a technologically good choice, it’s an essential step to take so you can realistically negotiate. Otherwise every time your contract renewal comes around, you’re weighing up the cost of the price hikes with the insane cost of migration. OpenTelemetry is the antidote to corporate ransomware in the Observability space.

50

3 Comments

Making the complex simple across multi cloud. Quick vid to show the power of the makers of Terraform. Terraform stacks is a new approach to coordinating the provisioning and management of dependent Terraform configurations. Environments just got alot easy. COPYPASTA just got better 🤙 #terraform

13

1 Comment

We're spending way too much time thinking about if something is running on Kubernetes. Why? Because Kubernetes isn't easy. When you're setting up a truly distributed, highly available environment with Kubernetes, it's not an easy task. Is it easier than bare-metal and ESXi? Yes. Does that mean it's easy? No. There's still a MASSIVE amount of confusion around Kubernetes (even from myself). And that's the problem. We need to stop putting so much effort into getting "really good" at Kubernetes and instead think about getting really good at distributing software in a highly performant fashion. We're spending way too much time managing underlying platforms. #kubernetes #devops #platformengineering

26

Here's a very simple example of the power of the #MQSCX =WHERE clause. It's like the #IBMMQ MQSC WHERE clause but with most of the restrictions removed. #WorkSmartWithMQ

9

"Our remote BuildKite-based CI takes ~5m30s to verify a code change is ready to ship for HEY. My Intel 14900K-based Linux box can do that in less than half the time. So we're going to drop the remote runners and just bring CI back to developer machines." I don't agree with DHH about everything - but this I kind of do. I have actually been waiting for containers/Kubernetes, coupled with just how amazingly powerful a $5k desktop or Macbook Pro is now, to bring back much of the dev pipeline/lifecycle to the devs' machine(s). Containers can now easily make the build environment consistent and reproducible between machines. And Kubernetes can even make the foundational IaC for spinning up the whole app/environment run locally just as in production for local testing. And much of the CI/CD tooling has been made to run on K8s - and so can run just as well locally on a local K8s on the dev machine now too. If you are a huge enterprise or a bank sure - the risk of doing CI/CD on a less trusted dev machine likely isn't worth it. But, for most companies, I think pulling it all back to local and getting 3-4 years deprecation at a known and fixed cost out of a high-end developer rig that is doing "all the things" other than running prod is starting to look very attractive (both for cost and performance/productivity reasons). Plus I imagine it will also make developers happier to both get these better machines as well as more attention/help to get their local dev environments set up quickly and properly on them too...

9

3 Comments

It's 5:47 PM PT, do you know where your geekette is? Wrapping up the K8s cluster creation/cleanup, PriorityClass, Namespace, Deployments, Service, Horizontal Pod Autoscaling (HPA), Grafana K6 and observing lower-priority deployments/pods being evicted when additional higher-priority deployments/pods are spun up to support an increased load on the associated high-priority service. Next steps? Write up the results...and maybe rerun the deployments using containerized Jupyter notebooks over the weekend? Kubeflow anyone? 🤖🤣🤓🐉

3