Atlas: Pod Stuck Terminating

What this usually means

Kubernetes is waiting for cleanup: a finalizer to complete, a volume to detach, or a node/kubelet to acknowledge termination. Force-deleting too early destroys evidence and can widen impact.

What to inspect first

Read deletion timestamps, finalizers, and node health.

• If the node is NotReady, termination may be blocked on kubelet reachability.
• Finalizers are explicit; identify which controller owns them.

kubectl get pod <pod> -n <ns> -o json | rg -n "deletionTimestamp|finalizers"

kubectl describe pod <pod> -n <ns>
kubectl get node -o wide

Resolution guidance

Prefer restoring controller/node health over forcing deletes.

• If a node is unhealthy, contain impact (cordon/drain posture) and recover/replace the node.
• If finalizers are stuck due to a failing controller, restore the controller and let it clean up.
• Force deletion is a last resort; when used, document and audit it—especially with stateful workloads.