Playbook

OOMKill Discovery

Diagnose and resolve Out-Of-Memory (OOM) kills in Kubernetes.

Check pod status for 'OOMKilled' reason in 'kubectl get pods'.

Inspect 'kubectl describe pod <name>' for exit code 137.

Look for 'Out of memory: Kill process' in node dmesg or syslog.

Run 'kubectl get events --field-selector reason=OOMKilling -A' for faster cross-namespace discovery.

Note the pod's QoS class: BestEffort pods are killed first, then Burstable, then Guaranteed.

Temporarily increase memory limits if headroom is available on the node.

Check for sudden spikes in traffic or data processing volume.

Verify if the application's heap size is correctly tuned for the container limit.

Distinguish request vs limit: requests set QoS class; limits trigger the OOM kill. Raising limits without raising requests does not change QoS class.

Optimize application memory usage (buffer sizes, concurrency).

Implement better memory management or switch to a more efficient runtime.

Use Vertical Pod Autoscaler (VPA) to right-size resource requests.

Set memory requests equal to limits for critical workloads to guarantee Guaranteed QoS class.

Find OOMKilled Pods

kubectl get pods -A | grep -i OOMKilled

Cross-namespace OOM events

kubectl get events --field-selector reason=OOMKilling -A

Check Node OOM Events

journalctl -k | grep -i "out of memory"

Get Pod Resource Usage vs Limits

kubectl top pod <pod-name> --containers

Inspect pod QoS class

kubectl get pod <name> -o jsonpath='{.status.qosClass}'

Need this managed for you, not just automated?

We're also a hands-on DevOps consultancy — Kubernetes, CI/CD, and cloud infrastructure.