kubernetes-job-troubleshooting

Check pod status and age with kubectl get pod, review current and previous logs, describe the pod for events, and check Job completions. A pod Running far longer than expected with repeating log output usually means a --wait command polling a hung external operation.

List ReplicaSets to find the working revision, then run kubectl rollout undo deploy/<name> --to-revision=<N>. The old ReplicaSet keeps serving during a broken rollout, so rollback restores service without building a new image first.

The pod is often a client calling an external system whose agent died mid-operation, leaving a stale lock. Agent status may report OK while actual work produces zero output, so check the external system's logs directly rather than trusting pod status.

No, --serviceaccount is not a valid flag for kubectl run. Use a Pod manifest applied with kubectl apply, or pass --overrides JSON, when you need a service account, secrets, or volume mounts for ad-hoc debug pods.

With the default concurrencyPolicy: Forbid, a stuck or failed Job blocks all future scheduled runs. Delete the stuck Job and set failedJobsHistoryLimit and successfulJobsHistoryLimit on the CronJob to prevent accumulation.

Two Certificate resources writing to the same Secret conflict: one auto-created by the Gateway's cert-manager.io/cluster-issuer annotation and one manual. Delete the manual Certificate, clear stale cert-manager annotations from the Secret, and let the Gateway-owned cert reconcile.