New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
delete PropagationPolicy and OverridePolicy at the same time, lead to workload recover to origin state #3873
Comments
Similar behavior also occurs during creation OverridePolicy and PropagationPolicy. |
Hi @wu0407 Thank you for your feedback. The deletion of When users delete I would like to understand why you need to delete |
I don't want deployment running on that cluster, but continue run on karmada cluster. Because karmada apiserver is reuese exist apiserver. relate issue: |
|
ResourceBinding currently has the same lifecycle as the resource template. If the resource template is not deleted, then ResourceBinding will also not be deleted.
I have created an issue to track this problem. #3886 |
@wu0407 @chaunceyjiang Thanks~, Your discussion has provided me with a new perspective. In fact, this issue has been discussed before, #3681 One previous conclusion was:
I think now we can further optimize this conclusion. The lifecycle of rb and resource templates remains consistent, and users can decide whether to clean up the scheduling results on rb with pp information so that they will no longer be synchronized. What do you think? |
if users choose to clean up the previous scheduling results, should the resource template be matched by other new appropriate policies? I think cleanup conflicts with rescheduling. it would be great if you could explain it to me. |
I don't think these are scheduling results. I believe they are scheduling information. Since the PP has already been deleted, the related information should also be cleaned up. |
I think that regardless of how users choose, it should be allowed for resource templates to match with new Propagation Policies. This point may not conflict.
When clearing the resource template of pp matching information, it is necessary to clean up the information on rb. |
I agree with you. The statement about the
Do you mean to clean directly without requiring user selection? |
Yes. |
I tend to agree a little, but I'm not sure if there are users who rely on this. It might be a good idea to hear what others think at the community meeting. |
/cc @RainbowMango |
I would recommend turning on a featuregate as a first step, the default behavior remains the same as it is now. Open the featuregate if the user want clean directly. Then someday in the future, when everything is stable, the default behavior can be modified and cleanup this featuregate. |
A simple version evolution example: Just let the featuregate open by default and do not need to cleanup. If a direct cleaning solution is adopted in the end, I suggest notifying users in advance in 1.7, and execute in 1.8 |
In another issues #3681 , we've discussed it yet, like "I prefer to add a new gate in pp to determine this behavior, and the default is consistent with the current one to be compatible with the existing data. User can open new gate in pp to change this behavior if he knows what he is doing."
For compatibility reasons, I recommend using the current behavior as the default. But if we can do some research, and the research results show that users prefer to clean up rb when deleting pp, then I suggest that we can adopt the method that @liangyuanpeng said, just like k8s' strategy for new featuregate. |
By the way, there will be new featuregate in pp that seems better than a global featuregate switch, casue user can control the action for each pp. I just leave the problem here that how's the default action going. cc @olderTaoist |
I'm looking and thinking about the use case:
I'm asking a further question, why can't remove it even in case of reusing the API server? Feel that it'd be better to have a chat at the community meeting. |
Hello everybody, @wu0407 @chaunceyjiang @liangyuanpeng @whitewindmills @weilaaa @olderTaoist Can we discuss this issue at the community meeting next Tuesday (2023-08-15)? I hope that we can reach some consensus and make progress towards resolving this issue during the meeting. |
What happened:
I want to clean the member resource, so I delete
OverridePolicy
andPropagationPolicy
on the same time.The OverridePolicy is to Override deployment spec.replicas to 0.
Many pod create by deployment controller on member cluster after delete action.
rule.yaml:
What you expected to happen:
no pod be created and deployment is deleted.
How to reproduce it (as minimally and precisely as possible):
Anything else we need to know?:
Environment:
kubectl-karmada version
orkarmadactl version
):The text was updated successfully, but these errors were encountered: