Find which pod is OOMKilling and propose a fix
Wann einsetzen: You're paged for memory pressure and don't have a dashboard up.
Voraussetzungen
- Server/skill installed and authenticated — See repo README
Ablauf
-
Spot the OOMsFind pods in any namespace with restartCount > 3 in the last hour. Show pod, namespace, last termination reason.✓ Kopiert→ Targeted list, likely with OOMKilled in the reason
-
Propose the fixGet the limits/requests for the worst offender. Tell me what to raise and write an apply_manifest dry-run patch.✓ Kopiert→ Resource diff + a dry-run apply
Ergebnis: Targeted patch ready to apply, RBAC-checked.
Fallstricke
- MKP only exposes what the bound SA can do — MKP only exposes what the bound SA can do — to enable apply_manifest you must bind an SA with patch permission.