TODO List: Enhance Cluster Lifecycle Management

[Garrybest]

What would you like to be added: According to https://github.com/karmada-io/karmada/issues/1640#issuecomment-1119377312, I'd like to enhance cluster lifecycle management. Here is the todo-list.

• [x] Add a taint for the clusters which are unhealthy for a period of time.#1781
• [x] Add taint manager. #1945
• [ ] Remove cluster scheduling result in spec.clusters when the cluster is terminating. #2373
• [ ] Remove Failover from scheduler.

Why is this needed: Move some functions away from scheduler and make the scheduler more concise.

[RainbowMango]

Hi @Garrybest I guess the task(Remove cluster scheduling result in spec.clusters when the cluster is terminating.) is included in #1945.

Should we update the task list?

[Garrybest]

Hi @Garrybest I guess the task(Remove cluster scheduling result in spec.clusters when the cluster is terminating.) is included in #1945.

Should we update the task list?

Well, I'm afraid not. This task is to drain a cluster by using NoExecute taint when we decide to unjoin it, while #1945 focuses on how taint manager works.

[Garrybest]

I have an idea. Maybe we could implement this in cmd karmadactl drain. Just like what kubectl drain does, when we want to unjoin a cluster, we cordon this cluster first and then remove RB's cluster scheduling result in spec.clusters. At last, we use karmadactl unjoin.

[XiShanYongYe-Chang]

/assign @Garrybest

[RainbowMango]

Great thanks to @Garrybest !!!