Some questions about VPA

[xiazhan]

Hi @jbartosik or someone else working on VPA

This is Allan from Microsoft who attended your weekly meeting last week. Thanks a lot for answering our questions on the meeting.
Now I have some extra questions:

• VPA have 3 parts: Recommender, Updater and Admission Controller. Do we have plan to support HA such that we can have multiple replicas for Recommender, Updater and Admission Controllers? Or we already support HA.
• For security patch, would we able to get it before announcement? If so, is it okay for us to release in public via our image repo once we get it?

[jbartosik]

Hi,
currently all components of VPA (Recommender, Updater, and Admission Controller) assume there is only 1 instance of each running[1]. I currently don't have capacity to implement HA. I'm happy to discuss design, review PRs and accept contributions implementing it.

For security patch do you mean fix for #5455? We're doing our best to release it quickly but I don't know when exactly we will release it.

[1] You can run multiple instances of for example recommender and tell them to watch pods in different namespaces (using this flag) or run multiple recommenders with different names (alternative recommenders) but those allow you to reduce load on a single instance / run different configurations and are not suitable to keep extra instances around to improve availability.

[gouglhupf]

Hi,
I have a followup question for the HA. I can see that the recommender and updater should be the only instances communicating with K8s API / modifying resources. What would prevent the admission controller to run with multiple pods? Does it store some local state?

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[gozer]

/remove-lifecycle stale

[voelzmo]

@gouglhupf You can run multiple instances of the admission-controller, we also do this in our platform. There's no local state and this works just fine.

@gozer was this also the answer you were looking for, or is there anything else left in this issue that you were wondering about?

[gozer]

I was hoping for at least leader elections for vpa-updater and vpa-recommender

[voelzmo]

As stated above: this is currently not supported. My understanding is that you're trying to raise a feature request, in that case I'd ask you to open a separate issue, so we can split out a dedicated discussion about that and close this issue, which was more about a few generic questions. Thanks!

[Shubham82]

As stated above: this is currently not supported. My understanding is that you're trying to raise a feature request, in that case I'd ask you to open a separate issue, so we can split out a dedicated discussion about that and close this issue, which was more about a few generic questions. Thanks!

Closing this issue as per the above comment.

[Shubham82]

/close

[k8s-ci-robot]

@Shubham82: Closing this issue.

In response to this:

/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.