[k8s API machinery] Exception Request for KEP-2340

[Marek Siarkowicz]

Hello!

I would like to request an exception for KEP for 1.31:

• Enhancement name: Consistent List From Cache
• Enhancement status (alpha/beta/stable): Alpha -> Beta
• SIG: api-machinery
• k/enhancements repo issue #: https://github.com/kubernetes/enhancements/issues/2340
• PR #’s: https://github.com/kubernetes/kubernetes/pull/126329
• Additional time needed (in days): None. All work completed and approved by @wojtekt, we just need the exception to be approved.
• Reason this enhancement is critical for this milestone: 
  The "Consistent List from Cache" enhancement addresses a fundamental performance bottleneck in Kubernetes. As detailed in this blog post: https://github.com/kubernetes/website/blob/fe8996a5accced1441085c84b97663b23c5c8c13/content/en/blog/_posts/2024-06-21-consistent-rea especially under high load especially under high loadd-from-cache.md, it significantly improves the latency and efficiency of LIST requests, reducing the latency up to 3x and CPU usage by 30%. This improvement is vital for maintaining the scalability and reliability of Kubernetes clusters.
  
• Risks from adding code late: (to k8s stability, testing, etc.):
  The primary risk lies in modifying the conversion webhook for parallelization so close to code freeze. While the core "consistent list" functionality has been extensively tested in K8s CI and scalability environments, the webhook change is a late addition. Thorough testing of this change is essential to mitigate any potential negative impact on stability.
  
• Risks from cutting enhancement: Removing this enhancement would have a severe impact on Kubernetes performance. When the feature was temporarily disabled, scalability tests showed an 8x regression in latency. Without this change, Kubernetes v1.31 would regress to performance levels 2x worse than the previous release. This demonstrates that the enhancement is crucial for large-scale Kubernetes deployments.
  
  
  
  Thanks,
  Marek