Issue

Deleting a CRD hangs indefinitely in Terminating and never completes.

Explanation

Kubernetes won’t fully remove an object that has finalizers — instead, it sets a deletionTimestamp and waits. The CRD gets unstuck only once its finalizers are cleared.

The customresourcecleanup.apiextensions.k8s.io finalizer tries to clean up all CR instances before the CRD goes away, and does so by calling the operator’s admission webhook. If the webhook is gone (e.g. after a partial uninstall), those calls fail, the finalizer never completes, and the CRD is stuck. To make things worse, once the CRD enters InstanceDeletionInProgress, the API server blocks writes to its CRs — so you can’t remove the finalizer from the CRs either.

Resolution

Forcibly remove the finalizer from the CRD:

kubectl patch crd <your-crd-name> \
  --type=json \
  -p='[{"op": "remove", "path": "/metadata/finalizers"}]'

This unblocks the pending deletion and the CRD will disappear shortly after.

This skips the normal CR cleanup. Any existing CR instances may linger in etcd, which can cause unexpected behavior if the CRD is reinstalled later. Check for strays with kubectl get <custom-resource> -A after the CRD is gone.