Init:CrashLoopBackOff
Init containers run one at a time before the app containers, and each must exit 0. Init:CrashLoopBackOff means one of them keeps failing. Init:Error is the same failure before the back-off starts. Init:0/2 means none of two init containers has completed yet.
Diagnose in order
Section titled “Diagnose in order”- Which init container?
kubectl get pod <pod> -o jsonpath='{range .status.initContainerStatuses[*]}{.name}: {.state}{"\n"}{end}'. - Its logs, including the previous attempt:
kubectl logs <pod> -c <init-container> --previous. - Its exit code: in
kubectl describe pod, under Init Containers, look atLast State.
Usual causes
Section titled “Usual causes”| Init container does | Usual failure |
|---|---|
Wait for a dependency (until nc -z db 5432) |
The dependency’s Service has no endpoints, or a NetworkPolicy blocks it |
| Run database migrations | The migration fails, or two pods run it concurrently and one loses a lock |
| Fetch config or secrets | Missing credentials or no network route to the secret store |
| Fix file permissions | Pod Security forbids running as root, so chown fails |
Fix the cause the logs show. Two design rules prevent most of these: don’t run migrations from an init container in a Deployment with many replicas (use a Job), and give “wait for dependency” loops a timeout and a clear log line so the failure is visible.