Tutorials
Investigate a restarting Pod
Find a deliberate crash, read its previous logs, and identify the command that caused it. No AI provider is needed.
Before you begin
Install Plous and prepare a disposable cluster with the playground requirements.
Run shell commands from the Plous checkout. Examples use context minikube . Replace that name consistently if yours differs.
Setup and cleanup affect the playground namespaces, including their PVCs. Use a cluster reserved for these examples.
1. Create the example
sh
python3 scripts/playground.py plan --context minikube --failures crash python3 scripts/playground.py setup --context minikube --failures crash
The failing Pod is broken-crash , with container example in namespace plous-failures . Wait for its image to download and the container to exit.
2. Find the Pod in Plous
sh
plous --data-dir /tmp/plous-pod-tutorial
Use :K to select minikube and check the header.
Press 22 , then type :pods and Enter.
Press n and choose plous-failures .
Press / , type broken-crash , and Enter.
Open the Pod with Enter.
The restart count should grow. Depending on timing, the container may be running, terminated, or waiting in CrashLoopBackOff. Use :g after another exit if needed.
An image-pull or scheduling error means the example has not reached the intended crash yet.
3. Follow status, events, and logs
In Describe, check the container's current and last termination state. It is designed to exit with code 1 .
Press e for Events. Look for repeated starts and backoff warnings.
Press l for logs, then p for the previous instance. Use o to choose example if needed. Look for:
text
ERROR deliberate crash; previous logs should retain this
Previous logs require a terminated instance. If unavailable, wait for another exit and reload with g , or read the current instance while it runs.
4. Check the cause
Press Esc, then y for YAML. Find the example container's command. The fixture prints the message, sleeps, and exits:
sh
echo 'ERROR deliberate crash; previous logs should retain this' sleep 3 exit 1
The nonzero exit causes the restart. CrashLoopBackOff is Kubernetes's retry delay, not the underlying cause.
No edit is needed for this exercise. In a real workload, inspect its owning controller before changing configuration. Restarting the same failing command will not fix it.
5. Clean up
Exit Plous with :quit , then run:
sh
python3 scripts/playground.py cleanup --context minikube
This removes both healthy and broken playground examples. It keeps the cluster, kubeconfig, and local Plous profile. See cleanup limits.
Next steps
Try other failure cases or investigate with plous-debug.