[SIG-compute] Meeting Notes 2026-04-15
Andrew Burden
2026-04-15 MINUTES
Recording: https://youtu.be/qbNx6bHu5LY
Attendees: Itamar Holder, Jan Schintag, Michal Skrivanek,Barak Mordehai Orel Misan
Introductions: Welcome everyone to the KubeVirt weekly sig-compute meeting
Do we have any new members this week that would like to introduce themselves?
Itamar Holder General note: Important emails for VEP owners:
Agenda and Notes:
Michal Skrivanek Talk about APIs suggested w.r.t. The migration improvement VEPs
Question: can we remove alpha fields from a v1 API group?
Michal Skrivanek: AFAICT the pruning is enabled by default and removed fields would be permanently removed from etcd.
Open Floor:
VEP Tracker:
Pull Requests that need attention: https://github.com/kubevirt/kubevirt/pulls?q=is%3Aopen+is%3Apr+label%3Asig%2Fcompute
Bug/issue scrub: https://github.com/kubevirt/kubevirt/issues?q=is%3Aopen%20is%3Aissue%20label%3Asig%2Fcompute%20-label%3Akind%2Fflake
(/triage {accepted | build-watcher | duplicate | needs-information | not-reproducible | unresolved})
Flake issue/PR scrub:
(/triage {accepted | build-watcher | duplicate | needs-information | not-reproducible | unresolved})
Zoom chat:
00:02:02 Itamar Holder (@iholder101): https://docs.google.com/document/d/1VTeHa-QxppNcDQbpoddrucBz1vXjxWB0JLARfS5Uuqs
00:07:58 Jed Lejosne: Lubo's point still stands, alpha feature gate does that
00:08:10 xpivarc: Reacted to "Lubo's point still..." with 👍
00:08:47 Itamar Holder (@iholder101): Yes, I agree.
Usually though we do at least try to come up with the right API from alpha. In here we sort of intentionally expose controvercial API from the start.
00:08:51 Itamar Holder (@iholder101): But I think that's fine!
00:09:00 Itamar Holder (@iholder101): And alpha is a great solution to experiment.
00:14:42 xpivarc: We can talk about this in VEP, one or more fgs
00:18:31 alayp: If you want to deal with ENV variables/ structured data inside of virt-launcher pod, look at the device metadata API: https://github.com/kubernetes/enhancements/issues/5304
00:23:11 Felix: You can delete separate alpha API groups easily
00:23:20 Felix: but alpha features that land in the main v1 API are hard to delete
00:24:01 Itamar Holder (@iholder101): I didn't understand it that way. Maybe maybe you're right.
Either way, we need to be sure about that - that's pretty crucial.
00:24:34 Felix: well, technically we can delete anything from v1 again
00:24:42 Vladik Romanovsky: yes
00:24:44 Felix: but I'd say it's weird for a stable api to change constantly
00:25:04 Vladik Romanovsky: we are talking about alpha only
00:25:14 Felix: separate group?
00:26:45 Vladik Romanovsky: I mean, no, just any api fields
00:26:45 Felix: We got that weird mix of a v1 API group and features still being held behind alpha feature gates
00:27:02 ehaas: How about you you just introduce a reserved annotation for "experimental" things?
00:27:19 Itamar Holder (@iholder101): @felix, So what happens when k8s want to add a field to the pod API under alpha?..
00:27:22 lyarwood: Does that happen after running kube-storage-version-migrator?
00:27:33 Itamar Holder (@iholder101): e.g. DRA?
00:27:54 Felix: I'd say adding is fine, but taking away?
00:28:21 Felix: I'm not strictly against removing, it's just that "v1" in an api group signals this won't go away to me
00:28:44 Itamar Holder (@iholder101): So what happens in k8s?
They added new fields for DRA, right? What if under alpha they decide to delete these fields or change their type? What happens then?
00:28:57 Itamar Holder (@iholder101): I mean, the fields were added to the pod API
00:28:59 Itamar Holder (@iholder101): which is v1
00:29:14 Felix: yeah, I really don't know
00:29:21 xpivarc: Kubernetes distinguish API/Field maturity afaiak
00:29:51 xpivarc: Say Pod will never go away, but alpa field in the Pod can go
00:30:13 Itamar Holder (@iholder101): I think we HAVE to be confident of that. That's really important.
Maybe we can add a doc somewhere to capture the findings with proofs. Because it's not the first time we're discussing this :)
00:30:19 alayp: no fields were removed in DRA, everything was additive in pod spec IIRC.
00:30:40 Itamar Holder (@iholder101): @Alay - right, but what if under alpha they would want to remove them?
00:31:05 Felix: It's not super clean to me, when viewing v1 Pod as the API contract... but I don't think anyone wants to go through the hassle of constants k8s api bumps either
00:32:18 lyarwood: https://github.com/kubernetes-sigs/kube-storage-version-migrator