Pod Stuck in Terminating — 4 Root Causes and How to Fix Each

A pod stuck in Terminating for minutes is one of the most reached-for --force --grace-period=0 situations in Kubernetes — and force is almost always the wrong fix. It hides the cause and can leave orphaned resources. Here are the four real causes.

Cause 1 — Finalizers

A finalizer is a key that blocks deletion until some controller does cleanup. If that controller is gone or broken, the pod hangs forever:

eknatha@prod ~

$ kubectl get pod stuck-pod -o jsonpath='{.metadata.finalizers}'

["example.com/cleanup"]

If the owning controller is genuinely dead, remove the finalizer (carefully):

eknatha@prod ~

$ kubectl patch pod stuck-pod -p '{"metadata":{"finalizers":null}}'

Cause 2 — Volume detach hanging

A pod with a PVC can't terminate until its volume detaches. If the node is unreachable or the storage backend is slow, it stalls:

eknatha@prod ~

$ kubectl describe pod stuck-pod | grep -A3 Events

Warning FailedUnmount ... timed out waiting for volume to detach

The fix is at the storage/node layer, not the pod — confirm the node is healthy and the CSI driver is running.

Cause 3 — A long preStop hook or grace period

eknatha@prod ~

$ kubectl get pod stuck-pod -o jsonpath='{.spec.terminationGracePeriodSeconds}'

300

A 5-minute grace period plus a slow preStop hook means the pod is correctly waiting. It's not stuck — it's draining. Check the hook before assuming a bug.

Cause 4 — A failing admission webhook

A validating/mutating webhook that times out can block the delete API call itself:

eknatha@prod ~

$ kubectl get validatingwebhookconfigurations

$ kubectl describe ... # check failurePolicy: Fail

If a webhook with failurePolicy: Fail is unreachable, every mutation — including deletes — blocks. Fix or temporarily relax the webhook.

Reach for --force and you'll delete the API object while the real pod and its volumes linger on the node. Find the cause first.