Kubernetes SIG API Machinery 20260304

Описание: [damdo] Plumbing of TLS Curves in SecureServingOptions
Following the conversation on https://kubernetes.slack.com/archives... I'd like to discuss next steps for this
Related PR: https://github.com/kubernetes/kuberne...

[rst0git] KEP-5823: Pod-Level Checkpoint/Restore
https://github.com/kubernetes/enhance...
List of questions for the meeting: [sig-api-machinery] Pod-level Checkpoint/Restore Subresources

[serathius] Consistent read from cache fallback, works great until defrag.
Defrag stalls member, watch gets delayed more than 3s, watch cache falls back reads to etcd.
Due to lack of APF protection (not aware of fallback), memory jumps (2GB pods)
Etcd 17GB-180GB
API server 50GB-160GB
Possible fixes:
Fallback down to APF and recalculate request cost. Should increase APF cost from 10 to 100, 10x less requests passing to etcd. Still memory jumping 2x. (1.36)
Etcd range streaming (proposal RangeStream Design, etcd 3.7, maybe K8s 1.37-1.38?).
Stabilizes etcd memory, but still API server manages whole LIST object.
Graceful degrade and return 429 status code instead of making things worse (https://tinyurl.com/k8s-graceful-shut... )

[ibihim|Krzysztof] KEP-3926 Corrupt Object Deletion, looking for support for review:
KEP-3926: Add a test to verify list error aggregation
Reviewer sttts, maybe (not a clear commitment)? Maybe someone could help review this?
KEP-3926: implement dry-run support for unsafe corrupt object deletion
Looking for an additional reviewer
Some comments but somehow stale
KEP-3926: test the delete handler with unsafe deletion flow
Reviewer jpbetz
kep-3926: add integration tests for CR and bit-flip corrupt object deletion
Reviewer jpbetz