[Exception Request] KEP-4680 DRAResourceHealth v1beta1 kubelet API, code freeze, v1.37

30 views
Skip to first unread message

Harshal Patil

unread,
Jul 21, 2026, 12:43:53 PM (11 days ago) Jul 21
to kubernete...@googlegroups.com, releas...@kubernetes.io, kubernetes-...@googlegroups.com
Hello Release Team,

I request a code freeze exception for the v1.37 release cycle.

Enhancement name: KEP-4680: Add Resource Health Status to the Pod
Status for Device Plugin and DRA

Enhancement status (alpha/beta/stable): beta

SIG: Node (WG Device Management)

k/enhancements repo issue #:
https://github.com/kubernetes/enhancements/issues/4680

PR #'s: https://github.com/kubernetes/kubernetes/pull/139477

Additional time needed : 5 days because API reviewers' bandwidth is
limited near code freeze time.

Reason this enhancement is critical for this milestone: The PR
promotes the DRAResourceHealth kubelet gRPC API from v1alpha1 to
v1beta1, matching the feature's beta graduation in this cycle. Without
it, the beta feature ships with an alpha only API for another release.

Risks from adding code late: Low. The PR is complete, has been
reviewed and approved by Patrick and Ed, and is passing CI. The only
outstanding item is API approver sign-off for
staging/src/k8s.io/kubelet/pkg/apis, where the v1beta1 proto is
identical to the existing v1alpha1, so there is no change in API
semantics to review. The change is internal to the kubelet and its
plugin interface rather than a REST API, and it was verified end to
end against real drivers on both the v1alpha1 and v1beta1 paths.

Risks from cutting enhancement: The PR also introduces the new
WatchHealthStatus method on the DRAPlugin helper interface, which
moves gRPC serving into the kubeletplugin helper so that future API
promotions never require driver changes again. If the PR slips to
1.38, drivers adopting health reporting in the 1.37 timeframe must
hand implement the v1alpha1 gRPC service and then migrate to the new
interface one release later. Landing it now means driver authors only
ever write against the final interface.

Thanks,
Harshal Patil

Harshal Patil

unread,
Jul 21, 2026, 12:53:59 PM (11 days ago) Jul 21
to sig-...@kubernetes.io, releas...@kubernetes.io, kubernetes-...@googlegroups.com

Adding sig-...@kubernetes.io, since the old googlegroups address for SIG Node bounced. Full request restated below for the SIG Node list.

Mrunal

unread,
Jul 21, 2026, 9:06:51 PM (10 days ago) Jul 21
to Harshal Patil, sig-...@kubernetes.io, releas...@kubernetes.io, kubernetes-...@googlegroups.com
+1 from sig-node.

--
You received this message because you are subscribed to the Google Groups "kubernetes-sig-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-re...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/kubernetes-sig-release/CAHY10_khpara%3DPebM3JbzD9Ccs2t4iayr1-muCFaRaPQaZQX1A%40mail.gmail.com.
Reply all
Reply to author
Forward
0 new messages