When a Kubernetes container crashes immediately or soon after starting, Kubernetes does not restart it continuously. It waits between attempts, increasing the delay each time to reduce wasted system resources. section 3.2.3
A failing container
A container might restart after 10 seconds, then 20 seconds, then 40 seconds. After repeated failures, its status becomes CrashLoopBackOff, signaling a problem that needs attention rather than a normal operating condition. section 3.2.3
Investigate repeated failures by:
- Checking pod events with kubectl describe pod $POD_NAME. section 3.2.3
- Reading current logs with kubectl logs and previous-instance logs with kubectl logs -p $POD_NAME. section 3.2.3
- Verifying external dependencies, such as whether a database is running and reachable. section 3.2.3
