OOMKilled means a container used more memory than it was allowed and Kubernetes stopped it. The signs are the same every time. A pod is a running copy of an app.
Know the signs
The container’s last state shows reason OOMKilled and exit code 137.
Read the limit and the usage
kubectl describe pod shows the last state and the memory limit. kubectl top pod shows what the pod uses now.
kubectl describe pod <pod> -n <ns>
kubectl top pod <pod> -n <ns>kubectl top needs metrics-server in the cluster.
Which limit stopped it
A container is stopped when it passes its own memory limit. A pod can also be stopped when the node itself runs out of memory. The last state of the container and the events on the node say which one happened.
Fix it
Pick one of two moves. Raise the limit to fit the use you observed, or fix the allocation the logs point to.
kubectl set resources deployment/<name> -n <ns> -c <container> --limits=memory=512MiThen confirm the restart count stops rising.
What kwatch sends
For an OOM kill, kwatch sends one message with the cause and, with Pro, the command to run. It closes it once when the pod is healthy again.