kubectl debug: From Live Pods to Crash Loops
Inspect running and distroless workloads, compare debug copies, and diagnose a delayed crash loop in a disposable K3s cluster. Finish with a read-only worker-node inspection.

A missing process in ps, a missing shell, and a restarting worker can look similar. In this gym you'll distinguish them using evidence, without restarting the original workload yourself.
Work on dev-machine. gym-k invokes the real kubectl with the exercise namespace and identity selected. For example, gym-k get pods shows the Pods for the current drill. The app fixtures and records are synthetic; don't use a production kubeconfig.
- Inspect a running process and compare an untargeted toolbox with
--target. - Check a distroless application and repeat a process check with a restricted profile.
- Change a command or image in a debug copy, then compare shared and isolated process views.
- Diagnose a delayed crash, reproduce it, and test a controlled change.
- As an optional extension, identify the worker OS and clean up the node debugger.
Practice in a real terminal
Use a regular shell on a real Linux VM. Each rep is marked complete automatically when you reach the goal.
One skill at a time
Each module focuses on one command or concept, with short exercises to help you get comfortable using it.
Repeat until it sticks
Revisit the gym after 1, 3, 7, 14, 30 days to build command-line habits and remember what you've learned.
Part of your Daily Practice
The gym appears in Daily Practice when it's time to repeat it. Pause when you need a break, or mark it complete when you're comfortable with the skill.
See your progress
Review past sessions to see how long you practiced and how many reps you completed.
Help when you get stuck
Use hints and feedback based on what's happening on your VM to work out the next step without being given the answer.