Open
Description
The e2e tests used to fail due to CRDs not being deleted in a timely manner. The CRDs would be in a terminating state with a specific error message. Eventually, after a few minutes, the CRDs are deleted.
This is handled in our e2e's by a 5-minute timeout while checking for CRD deletion.
This was the interesting condition on the CRDs:
{Terminating True 2024-11-06 21:00:31 +0000 UTC InstanceDeletionFailed could not list instances: storage is (re)initializing}
References to this error:
- https://github.com/kubernetes/enhancements/pull/4557/files
- https://github.com/kubernetes/kubernetes/pull/124642/files
- Error starting operator with K8s v1.31.0 (429 Message: storage is (re)initializing) java-operator-sdk#2526
- https://github.com/kubernetes/kubernetes/blob/e2bf630940946df5bc161d224e4a9b2e191a3b2e/staging/src/k8s.io/apiextensions-apiserver/pkg/controller/finalizer/crd_finalizer.go#L199-L204
Originally posted by @tmshort in #1428 (comment)