Most people who use Kubernetes cannot debug it. They can deploy to it, because someone else wrote the manifests and the pipeline works. Then one day a pod goes Pending and stays there, and the entire team is lost.

This workshop fixes that. The method is that we are going to spend four hours breaking a cluster on purpose.

You will stand up a cluster on your laptop and then we will systematically ruin it. A pod that will not schedule, and you find out why from events instead of guessing. A liveness probe tuned just wrong enough to restart a healthy service forever. A ConfigMap change that silently does nothing. A resource limit that gets your process killed by the kernel with no useful error anywhere. A rolling deploy that hangs at 1 of 3 and never finishes. A DNS failure that looks like an application bug for twenty minutes.

Every one of these is something I have hit in production running clusters for a platform that tracks utility spend across 350,000 sites. Every one of them is boring once you know the failure signature. None of them are obvious the first time.
The fun part is that we build the broken things ourselves, which means you also end up fluent in Pods, Deployments, Services, ConfigMaps, Secrets, probes, resource requests and limits, taints and tolerations, and autoscaling. You just learn them from the failure side, which is the side that sticks.
Come in able to deploy. Leave able to diagnose.

What you will leave with:
• A repeatable first five minutes for any misbehaving cluster
• Fluency in kubectl describe, get events, and logs --previous, the three commands that solve most of it
• Six failure signatures you will recognize instantly next time
• Every manifest and every break, in a repo you keep, so you can run the drills again for your team