updating CRD causes connection loss with active watches of custom resources

[l1b0k]

What happened?

1. update CRD 's description fields
2. the client watching for the cr resource will lost watch Terminate custom resource watches when storage is destroyed #78029

What did you expect to happen?

As i only chang the CRD's descriptions , this is not necessory to notify all client about the change.
This increase signifient pressure for kube-apiserver in large cluster.

How can we reproduce it (as minimally and precisely as possible)?

update CRD fields, and you can see watch connection is lost

Anything else we need to know?

No response

Kubernetes version

$ kubectl version
Server Version: version.Info{Major:"1", Minor:"22+", GitVersion:"v1.22.15-aliyun.1", GitCommit:"707e514954f0f3ba8ce36face7cf7058403057bc", GitTreeState:"clean", BuildDate:"2022-09-22T03:45:47Z", GoVersion:"go1.16.15", Compiler:"gc", Platform:"linux/amd64"}

/sig api-machinery

[l1b0k]

/sig api-machinery

[Ritikaa96]

Maybe we can retitle it as:
updating CRD causes connection loss with active watches of custom resources.
for a better understanding.

[Ritikaa96]

/area api-server
/area custom-resources

[k8s-ci-robot]

@Ritikaa96: The label(s) area/api-server cannot be applied, because the repository doesn't have them.

In response to this:

/area api-server
/area custom-resources

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[Ritikaa96]

Whoops , my bad!
/area apiserver

[Sakuralbj]

Now crdHandler compare the spec and acceptedNames of crd to decide if crd update should be ignored.
So if some description info such as spec.versions.additionalPrinterColumns.description was changed,
kube-apiserver will destory storage and close watches. Can we make a more precise comparison here to reduce unnecessary connection close. In large-cluster the reconnection will bring huge load to apiserver. https://github.com/kubernetes/kubernetes/blob/master/staging/src/k8s.io/apiextensions-apiserver/pkg/apiserver/customresource_handler.go#L499

[Sakuralbj]

Maybe spec and acceptedNames here is not correct enough to compare against if a change is made on a CRD. Whether we can build a new struct to save necessary info which can be used to judge if storage need change.

[leilajal]

/triage accepted

[k8s-triage-robot]

This issue has not been updated in over 1 year, and should be re-triaged.

You can:

• Confirm that this issue is still relevant with /triage accepted (org members only)
• Close this issue with /close

For more details on the triage process, see https://www.kubernetes.dev/docs/guide/issue-triage/

/remove-triage accepted

[Ritikaa96]

hi @l1b0k does the error still exist on your side?

[jiahuif]

/triage accepted