Would like a descheduler that evicts pods have preferred node anti-affinity

[darrellschrag]

Is your feature request related to a problem? Please describe.
I typically try to schedule multiple replicas of pods across worker nodes. This helps reduce impact if a node is lost. I typically use preferred so that if I do lose a node, I will maintain my desired replica count. If a node is lost and no more are available, Kube will reschedule the pod on a node that already has one. When a new node is added back to the cluster, the descheduler does not evict one of the 2 pods on a single node so it can be rescheduled back to the new 3rd node.

  affinity:
    podAntiAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 100
        podAffinityTerm:
          labelSelector:
            matchExpressions:
            - key: app
              operator: In
              values:
              - app-name
          topologyKey: kubernetes.io/hostname

Describe the solution you'd like
My deployment spreads 3 pods across 3 nodes. I have the above preferred anti-affinity configuration. When a node is lost, a new pod is created on one of the 2 remaining nodes (preferred). When a 3rd node is reintroduced into the cluster, I want the descheduler to evict one of the 2 pods running on a single node so it can be rescheduled onto the new 3rd node.

Describe alternatives you've considered
Manually killing the pod.

What version of descheduler are you using?
descheduler version: 0.29

Additional context

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[k8s-triage-robot]

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle rotten
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten