Reading logs and events

Events explain what logs cannot

When a pod never starts, its logs are empty and the events tell the whole story: no node had capacity, the image could not be pulled, a volume would not mount, a probe kept failing.

For a pod that crashed and restarted, the current logs are from the new container. Fetching the previous container logs is how you see the actual crash.

Learn the status vocabulary. Pending is scheduling, image pull errors are registry or credentials, crash loops are your application, and an out-of-memory kill is a limits problem.