Make available HPA autoscaling API (v2beta2) in GKE 1.12 [135624588]

Fixed

Feature Request

Status Update

No update yet.

Description

ad...@google.com

created issue #1

Jun 19, 2019 08:20PM

Please describe your requested enhancement. Good feature requests will solve common problems or enable new use cases.

What you would like to accomplish: Please enable the HPA autoscaling API version V2beta2 in GKE 1.12.

How this might work: This is blocking [1] to work in GKE 1.12 since looks for the version v2beta2 API of HPA autoscaling.

[1]

https://github.com/zalando-incubator/kube-metrics-adapter

If applicable, reasons why alternative solutions are not sufficient: The adapter would not work if the right API version is available in the cluster.

Other information (workarounds you have tried, documentation consulted, etc): This is [2] a related open issue in Github

[2]

https://github.com/zalando-incubator/kube-metrics-adapter/issues/55

Comments

mi...@gmail.com <mi...@gmail.com> #2Jul 21, 2019 08:14AM

Any progress on this?

I just checked the latest GKE version (1.13.7-gke.8) and it's still not enabled :(

I'm a maintainer of the

https://github.com/zalando-incubator/kube-metrics-adapter project and would like to be able to support users of GKE. We could ofc. do changes on our side to accommodate this limitation in GKE, but I would appreciate if this can be fixed in GKE instead.

mi...@thoughtmachine.net <mi...@thoughtmachine.net> #3Sep 2, 2019 04:24PM

I'd be interested in this as well, could we make this happen?

gu...@gmail.com <gu...@gmail.com> #4Sep 4, 2019 10:32AM

Any updates?

ya...@geomagical.com <ya...@geomagical.com> #5Sep 4, 2019 04:44PM

Updates? this really feels like GCP is lagging behing a lot in terms of metrics!

mi...@metarouter.io <mi...@metarouter.io> #6Sep 10, 2019 07:53PM

Looking for an update here as well.

ro...@rewe-group.com <ro...@rewe-group.com> #7Sep 12, 2019 06:01PM

[Deleted User] <[Deleted User]> #8Sep 27, 2019 03:05PM

ha...@gmail.com <ha...@gmail.com> #9Oct 10, 2019 07:43PM

Any update on this? Why is GKE so far behind?

Message last modified on Oct 11, 2019 02:47PM

jo...@google.com <jo...@google.com> Oct 11, 2019 08:10AM

Assigned to jo...@google.com.

[Deleted User] <[Deleted User]> #10Oct 25, 2019 12:29AM

It looks like recent documentation mentions v2beta1 in its examples:

https://cloud.google.com/kubernetes-engine/docs/tutorials/custom-metrics-autoscaling

https://cloud.google.com/kubernetes-engine/docs/tutorials/external-metrics-autoscaling

gu...@gmail.com <gu...@gmail.com> #11Oct 25, 2019 08:53AM

#10: v2beta1 doesn't support taking into account multiple scaling metrics. So it is not super useful to be honest.

da...@we-riot.com <da...@we-riot.com> #12Oct 28, 2019 08:35AM

#11

https://godoc.org/k8s.io/api/autoscaling/v2beta1#HorizontalPodAutoscalerSpec but it does, the difference between v2beta1 and v2beta2 is in the ways you can describe the targets, example:

https://godoc.org/k8s.io/api/autoscaling/v2beta1#ExternalMetricSource

https://godoc.org/k8s.io/api/autoscaling/v2beta2#ExternalMetricSource

v2beta2 is a bit more descriptive and you can have AverageUtilization as well as Value and AverageValue

jo...@google.com <jo...@google.com> #13Nov 6, 2019 11:04AM

I've enabled the v2beta2 in the upcoming GKE 1.16 version (no timeline yet). I also backported the change to 1.15.5 which will become available after 1.15 goes GA, hopefully soon.

My recommendation is to wait for 1.15.5 to become available and upgrade if you need autoscaling v2beta2.

ch...@pluralsight.com <ch...@pluralsight.com> #14Jan 17, 2020 12:24AM

I just upgraded a cluster from 1.15.4 to 1.15.7 and can confirm the addition of autoscaling/v2beta2 (though I did not test it via autoscaler).

jo...@google.com <jo...@google.com> #15Jan 17, 2020 08:05AM

Marked as fixed.

Excellent!