kubernetes-retired/kube-aws

[0.14.3 => 0.15] Regression regarding Persistent Volume

sei-nicolas opened this issue · 3 comments

Using the same original cluster.yaml file:

  • With release 0.14.3, it's possible to use an Amazon EBS volume as a persistent volume, then declare a persitent volume claim to mount inside a deploy
  • With release 0.15, the pod creation fails with the error "Failed to get AWS Cloud Provider. GetCloudProvider returned instead"

I saw that this message could happen if kubelet is not started with the parameter "--cloud-provider=aws" but manually adding it (api/node) does not change anything.

Apologies, yes I hadn't realised that the PVS were no longer working with the cloud-controller-manager.

A fix is to add the kube-controller-manager flag --cloud-provider=aws which restores PVC functionality (but throws a deprecation warning on k8s 1.16).

I think the right solution will be to make cloud-controller-manager an experimental feature and also add in CSI ebs volume support so that they can be used together (which I believe is the intended path forward as the aws cloud provider is gradually removed from core k8s).

I'll look at a patch which we can release as v0.15.1 containing these fixes.

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

This issue was resolved with #1832.