OOMKilled Cheatsheet

kubectl describe pod <name> | grep -A 5 "Last State"

→ Shows reason (OOMKilled), exit code (137) and timestamp of the last crash.

kubectl get pod <name> -o jsonpath='{.status.containerStatuses[*].lastState.terminated.reason}'

→ One-line output, ideal for scripts and alerts.

kubectl get events -A --field-selector reason=OOMKilling | tail -20

→ Finds every OOMKill across the cluster, sorted by time.

kubectl top pod <name> --containers

→ Current memory per container - the most important live number.

kubectl top pods -A --sort-by=memory | head -20

→ Sorts every pod cluster-wide by memory consumption.

kubectl exec <pod> -- cat /sys/fs/cgroup/memory.peak

→ Maximum since container start - the only number that matters for right-sizing.

kubectl describe pod <name> | grep -A 3 "Limits\|Requests"

→ Shows what the manifest really says - often different from what you expect.

kubectl get pod <name> -o jsonpath='{range .spec.containers[*]}{.name}: {.resources}{"\n"}{end}'

→ Clean overview of every container in the pod.

kubectl describe node <node> | grep -A 5 "Allocated resources"

→ Total requests vs node capacity - reveals node pressure.

kubectl set resources deployment/<name> --limits=memory=1Gi --requests=memory=512Mi

→ Quick patch - for GitOps, commit it in the manifest.

kubectl rollout status deployment/<name> -w

→ Waits until new pods are stable, or surfaces errors.

kubectl top pod <name> --containers -w

→ Live monitor - usage under load should stay below the new limit.