Karmada not propagate resources to new member cluster

[liuchintao]

What happened:

I just want to propagate my deployment to all member clusters automatically, when they join.

apiVersion: policy.karmada.io/v1alpha1
kind: PropagationPolicy
metadata:
  name: nginx
spec:
  resourceSelectors:
  - apiVersion: apps/v1
    kind: Deployment
    name: nginx

After creating this PP, I find nginx deployment in cluster-1 and cluster-2.

But when I join a new member cluster-3, I find nginx deployment is not propagated to cluster-3

What you expected to happen:

Deployment should be propagated to new member cluster.

type Placement struct {
	// ClusterAffinity represents scheduling restrictions to a certain set of clusters.
	// If not set, any cluster can be scheduling candidate.
	// +optional
	ClusterAffinity *ClusterAffinity `json:"clusterAffinity,omitempty"`
}

How to reproduce it (as minimally and precisely as possible):

Anything else we need to know?:

Environment:

• Karmada version:
• kubectl-karmada or karmadactl version (the result of kubectl-karmada version or karmadactl version):
• Others:

[RainbowMango]

I just want to propagate my deployment to all member clusters automatically, when they join.

Yes, I think this is a reasonable use case. But now the scheduler doesn't re-schedule in case of cluster joining or removal.

@dddddai I can see there is TODO, do you mean to cover this case? https://github.com/karmada-io/karmada/blob/ed9b838056ee45720abf51b2d603548e6ce922bf/pkg/scheduler/scheduler.go#L384

[dddddai]

Yeah I was thinking so, there were similar issues https://github.com/karmada-io/karmada/issues/1644 https://github.com/karmada-io/karmada/issues/829#issuecomment-1107735202

I think the descheduler should be responsible for this case?

[RainbowMango]

I think the descheduler should be responsible for this case?

cc @Garrybest

I'll pay attention to this feature. @dddddai can you help to lead the effort?

[Garrybest]

Descheduler's duty is to evict replicas only. This issue focuses on scheduling a object to new joined clusters.

I think we could always s.scheduleResourceBinding(rb) when the placement is Duplicated. It is ok?

[liuchintao]

hello, is there anything new?

[RainbowMango]

@liuchintao we are working on it. @chaunceyjiang send a PR(#2301) for this. /assign @chaunceyjiang