Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Support in-place update of pod resources #15716

Open
skonto opened this issue Jan 21, 2025 · 3 comments
Open

Support in-place update of pod resources #15716

skonto opened this issue Jan 21, 2025 · 3 comments
Labels
area/autoscale kind/feature Well-understood/specified features, ready for coding.

Comments

@skonto
Copy link
Contributor

skonto commented Jan 21, 2025

/area autoscale
/area API

Describe the feature

  • Related KEP is going to be beta in 1.33.
  • We might want to think how this could be used with KPA. In the past we had a design document for VPA but the integration never landed. We probably want to revise that.
  • Some older example of the feature here.

cc @dprotaso @evankanderson @dsimansk

@skonto skonto added the kind/feature Well-understood/specified features, ready for coding. label Jan 21, 2025
@skonto skonto changed the title Support in place resource update for pods Support in-place update of pods resources Jan 21, 2025
@skonto skonto changed the title Support in-place update of pods resources Support in-place update of pod resources Jan 21, 2025
@evankanderson
Copy link
Member

We may want to also track the workload resize design sketch in the KEP: https://github.com/kubernetes/enhancements/blob/master/keps/sig-node/1287-in-place-update-pod-resources/README.md#design-sketch-workload-resource-resize

Without workload resize, we may need something like a custom VPA or running a VPA for each Pod.

@dprotaso
Copy link
Member

Container concurrency with tuned resource request/limits are really tied together and work well for horizontal scaling.

In place updates only really work when concurrency is very high (essentially you don't care to optimize the horizontal scaling case)

@skonto
Copy link
Contributor Author

skonto commented Jan 27, 2025

@dprotaso

In place updates only really work when concurrency is very high (essentially you don't care to optimize the horizontal scaling case)

I am thinking of scenarios where you may want to grow the pod instead of the number of instances to avoid for example cold starts e.g. AI cases. You may also not to do so from the beginning so you don't request a lot of resources depending on the traffic pattern e.g. bursty.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
area/autoscale kind/feature Well-understood/specified features, ready for coding.
Projects
None yet
Development

No branches or pull requests

3 participants