Need support for customized recommender in Vertical Pod Autoscaler (VPA)

[wangchen615]

Which component are you using?:
Vertical Pod Autoscaler

Is your feature request designed to solve a problem? If so describe the problem this feature should solve.:
Today, the current VPA recommends CPU/Mem requests and limits based on one recommender,
which recommends the future requests/limits based on the historical usage observed in a
rolling time window. As there is no universal recommendation policy that applies to all
types of workload, this feature requests supporting multiple customized recommenders in VPA.

Describe the solution you'd like.:
The support of a customized recommender can be implemented via a first-citizen approach. Namely, a dedicated field recommenderName can be added to the VPA object to indicate which recommender to use.

Describe any alternative solutions you've considered.:

• Develop a plugin framework for customizable recommenders.
• Develop a label selector approach.

Additional context.:
This issue is created to obtain a KEP number.

[bskiba]

Shouldn't we keep open for implementation? Or do you want to track separately?

[wangchen615]

Shouldn't we keep open for implementation? Or do you want to track separately?

Yes, let's keep it open until the implementation is merged.

[Jeffwan]

@wangchen615 What's the status of this VPA KEP? Is there a PR going on?

[wangchen615]

@Jeffwan , I am still working on the implementation. Will create the PR soon, in 2 weeks.

[k8s-triage-robot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle stale

[Jeffwan]

/remove-lifecycle stale

We are still working on this