Autoscaler: Allow scaling node group to/from 0.

[mwielgus]

This is an umbrella bug for allowing scaling some node groups to/from 0.

cc: @MaciekPytel @fgrzadkowski

[Shimi]

@mwielgus I was wondering if there is a special reason why this is enabled only for GCE and not for AWS or Azure?

[mwielgus]

@Shimi
The reason is quite simple. I'm a Google engineer and, for various reasons, I have limited access to AWS and Azure so I cannot write and test the scale-to-0 support on these cloud providers.

However I will happily accept the contributions from the community. It is just a matter of implementing:

https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler/cloudprovider/aws/aws_cloud_provider.go#L290

and

https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler/cloudprovider/azure/azure_cloud_provider.go#L238

Feel invited :).

[mwielgus]

cc: @mumoshu @andrewsykim

[mumoshu]

Thanks for informing me :)
I'm running production clusters on AWS and am interested in this feature.
I can submit a PR containing a POC but, as always, I need someone to help testing/reviewing/merging it.

[7chenko]

@mumoshu We're interested in this feature on AWS, so would be glad to test it on our deployment.

[mwielgus]

Closing in favor of #132.

[paolomainardi]

I don't understand if this feature can be activated also on GKE, if i try to set the minimn size to 0 i have the following error:

gcloud container clusters --project "my-prj"  update spark-op-services --enable-autoscaling --min-nodes=0 --max-nodes=10 --zone=europe-west1-b --node-pool=my-pool

ERROR: (gcloud.container.clusters.update) INVALID_ARGUMENT: bad node pool autoscaling data: cloud-kubernetes::INVALID_ARGUMENT: EXTERNAL: node_pool_autoscaling.min_node_count must be greater than 0..

[MaciekPytel]

This should work on GKE clusters using kubernetes 1.7. I don't think 1.7 clusters are available right now, but they should be very soon.

[paolomainardi]

Ok just tried now, master and pool upgraded to 1.7.0:

 ❯ gcloud container clusters --project "spark-int-cloud-services"  update spark-op-services --enable-autoscaling --min-nodes=0 --max-nodes=3 --zone=europe-west1-b --node-pool=gitlab-ci

ERROR: (gcloud.container.clusters.update) INVALID_ARGUMENT: bad node pool autoscaling data: cloud-kubernetes::INVALID_ARGUMENT: EXTERNAL: node_pool_autoscaling.min_node_count must be greater than 0..

[paolomainardi]

Kubectl version

 ❯ kubectl version                                                                                                                                              
Client Version: version.Info{Major:"1", Minor:"6", GitVersion:"v1.6.6", GitCommit:"7fa1c1756d8bc963f1a389f4a6937dc71f08ada2", GitTreeState:"clean", BuildDate:"2017-06-16T18:34:20Z", GoVersion:"go1.7.6", Compiler:"gc", Platform:"darwin/amd64"}
Server Version: version.Info{Major:"1", Minor:"7", GitVersion:"v1.7.0", GitCommit:"d3ada0119e776222f11ec7945e6d860061339aad", GitTreeState:"clean", BuildDate:"2017-06-29T22:55:19Z", GoVersion:"go1.8.3", Compiler:"gc", Platform:"linux/amd64"}

[paolomainardi]

Thread: https://groups.google.com/forum/#!topic/kubernetes-sig-apps/Npq2FDim43I