How to prevent deletion of resources for member clusters?

[chaunceyjiang]

Please provide an in-depth description of the question you have:

I used the following command by mistake. Now I want to delete coredns on the control plane, but I don’t want to delete the coredns of the member cluster. How should I do it?

karmadactl promote deployment coredns -n kube-system -C cluster1

What do you think about this question?:

Environment:

• Karmada version: v.1.4.0
• Kubernetes version:
• Others:

[RainbowMango]

We have a draft proposal for this:https://github.com/karmada-io/karmada/tree/master/docs/proposals/cleanup-propagated-resources

The proposal is not to solve the mistake but to provide a mechanism to control if the resource should be removed from member clusters.

[chaunceyjiang]

provide a mechanism to control if the resource should be removed from member clusters.

This document proposes a mechanism to specify that a member cluster resource should be removed from a managed cluster when leaving.

I don't want to remove the member cluster, I just want to delete the coredns of the control plane.

Because controlling coredns in the control plane is a dangerous operation.

[chaunceyjiang]

karmada/pkg/util/constants.go

Lines 12 to 18 in d161730

	// PropagationInstruction is used to mark a resource(like Work) propagation instruction.
	// Valid values includes:
	// - suppressed: indicates that the resource should not be propagated.
	//
	// Note: This instruction is intended to set on Work objects to indicate the Work should be ignored by
	// execution controller. The instruction maybe deprecated once we extend the Work API and no other scenario want this.
	PropagationInstruction = "propagation.karmada.io/instruction"

I think we can extend this instruction.

[sunbinnnnn]

I got the same case, and almost cause terrible disaster. my workaround is stopping the controller, and force delete RB, workload and WORK, then start the controller...

[chaunceyjiang]

and almost cause terrible disaster.

Yes

my workaround is stopping the controller, and force delete RB, workload and WORK, then start the controller...

This operation has caused me a lot of trouble.

[RainbowMango]

I'm trying to figure out which PR/Issue should be included in the coming v1.7 release which is planned at the end of this month.
I guess we don't have enough time for this feature, so I'm moving this to v1.8.

[XiShanYongYe-Chang]

Hi @chaunceyjiang Do you need to keep work at the same time in your use case?

Ref #4788 (comment)

[RainbowMango]

This issue could be resolved by #5577 (comment).

/close

[karmada-bot]

@RainbowMango: Closing this issue.

In response to this:

This issue could be resolved by #5577 (comment).

/close

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-sigs/prow repository.