This draft turns a storage incident note into a troubleshooting pattern: reproduce the issue with a minimal PVC and pod, inspect the events and Longhorn volume state, and use that evidence to separate application problems from storage attachment problems.
This draft captures a real troubleshooting pattern for 504 Gateway Timeout responses behind Kubernetes ingress by checking the ingress rule, backend service, pod health, and in-cluster connectivity from the ingress controller itself.
This draft captures a real ingress debugging flow on a bare-metal cluster: verify the service IP, inspect the ingress objects, deal with a Helm ownership conflict, reinstall the controller cleanly, and confirm that the new ingress class is actually serving the expected backend.
This is a carefully sanitized first-pass draft from a sensitive note. The useful pattern is still there: test the registry with curl against /v2/, vary the host and scheme deliberately, confirm the image naming path, and validate that the backup and retention tasks around the same service are in place.