Autoscaler: Undocumented minimum resource requests for VPA?

Created on 9 May 2019 · 10Comments · Source: kubernetes/autoscaler

I have a simple VPA on GKE that looks like this:

apiVersion: autoscaling.k8s.io/v1beta2
kind: VerticalPodAutoscaler
metadata:
  namespace: kube-system
  name: heapster-vpa
spec:
  targetRef:
    apiVersion: "extensions/v1beta1"
    kind: Deployment
    name: heapster-v1.6.1
  updatePolicy:
    updateMode: "Initial"

It results in scaled down cpu requirements for heapster but the Memory Requests jump to "87381333" for every container. Resulting in unneeded high memory requests for this pod.
I noticed the same behavior for other VPAs (eg. one for kube-dns).

Is there an undocumented minimum resource request and if yes, can it be changed?

lifecyclrotten

Source

fiws

👍2

Most helpful comment

looking for a way to do this as well.

/remove-lifecycle rotten

aranair on 12 Dec 2019

👍2

All 10 comments

There is a pod-recommendation-min-memory-mb flag which defaults to 250 MiB. This is a minimum recommendation for the whole Pod so if it has multiple containers, it will be split between these.

This flag cannot currently be changed in GKE.

kgolab on 9 May 2019

aw :( we are trying to lower resource reservation for our small dev cluster. And this was the only way I found to lower it for system deployments like heapster or kube-dns.

Can you recommend any workaround for this? Or will it be possible to adjust that default in the future? It seems quite high for small clusters.

fiws on 9 May 2019

👍2

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-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

fejta-bot on 7 Aug 2019

we still found no other workaround for this.

/remove-lifecycle stale

fiws on 7 Aug 2019

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-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

fejta-bot on 5 Nov 2019

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

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

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle rotten

fejta-bot on 5 Dec 2019

looking for a way to do this as well.

/remove-lifecycle rotten

aranair on 12 Dec 2019

👍2

lower it for system deployments like heapster or kube-dns

A workaround for this could be to set minAllowed = maxAllowed = desired. For example, setting it to 50 MiB should result in 50 MiB, even though pod-recommendation-min-memory-mb is higher.

amq on 18 Dec 2019

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/close

fejta-bot on 17 Jan 2020

@fejta-bot: Closing this issue.

In response to this:

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/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.