WIP: Allow kubelet to trigger rescheduling of pods

[ffromani]

Enhancement Description

In some cases the scheduler picks a node to run a workload, but once the kubelet tries admission it determines it cannot actually run the workload or it cannot run optimally. The simplest example is numa-aware scheduling in the current kubernetes version (without DRA/NRI) but this also applies to driver being removed or going unhealthy and so forth. This leads to runaway pod creation (kubernetes/kubernetes#84869).
This proposal aims to improve the scheduler/kubelet interaction to enable the workload to land on a optimal node.
It is additive to the usage of descheduler or any other external controller.

NOTE 20250611 SIG reviewers please be aware I don't have bandwidth to work on this atm, so this is safe to ignore until further notice. I needed an enhancement item for easier reference and tracking purposes.

• One-line enhancement description: enable the kubelet to trigger rescheduling of pods
• Kubernetes Enhancement Proposal: in progress
• Discussion Link:
  • v1 proposal (obsolete): https://docs.google.com/document/d/1-wJhiNy84w7tzFdo9HqwTu5DrVSuXFLGTUv8FBiRAAc/edit?tab=t.0
  • v2 proposal (current, WIP): https://docs.google.com/document/d/102ApSxz_9yUXeSkTaqcv5BViYSSdRv30A39b-rD09Po/edit?usp=sharing
• Primary contact (assignee): @ffromani
• Responsible SIGs:
  /sig node
  /sig scheduling
• Enhancement target (which target equals to which milestone):
  Alpha release target: TBD
  Beta release target: TBD
  Stable release target: TBD
• PRs by stage and milestone:
  • Alpha - v1.xx
    • KEP (k/enhancements) update PR(s):
    • Code (k/k) update PR(s):
    • Docs (k/website) update PR(s):

Please keep this description up to date. This will help the Enhancement Team to track the evolution of the enhancement efficiently.

[ffromani]

/sig node
/sig scheduling