Code Freeze Exception Request for KEP-5823

[Radostin Stoyanov]

Hello Release Team,

We would like to request a 5 calendar day code freeze exception for the following enhancement.

Enhancement name: KEP-5823 Pod-level Checkpoint/Restore

Enhancement status: alpha

SIG: Node (WG Checkpoint Restore)

k/enhancements issue: https://github.com/kubernetes/enhancements/issues/5823

PRs: https://github.com/kubernetes/kubernetes/pull/140366

Additional time needed: 5 calendar days

Reason this enhancement is critical for this milestone:

The Pod-level Checkpoint/Restore design has been under discussion since the first WG meeting in December 2025, went through several reviews from SIG Node, SIG API Machinery and SIG Architecture. This PR provides a foundation for reducing the cold-start time of AI inference workloads and enabling fault tolerance for long-running jobs (e.g., model training). Distributed inference frameworks such as NVIDIA Dynamo are already building out-of-tree workarounds to support this functionality. Introducing an off-by-default alpha in 1.37 establishes an in-tree mechanism that the ecosystem can converge on instead of each project implementing incompatible out-of-tree solutions.

Risks from adding code late:

The scope of this PR is limited to introducing only the CRI APIs that would allow implementing support for Pod-level Checkpoint/Restore in container runtimes (containerd and CRI-O).

Risks from cutting enhancement:

If this PR does not make the v1.37 release, it would delay the roadmap of the entire working group and many other KEPs that will begin building on Pod-level Checkpoint/Restore.

Thank you for your consideration,

Radostin Stoyanov