Use hostpath.csi.k8s.io across multiple node pools

[andymcmahan]

Hi.
When using csi-driver-host-path in GKE 1.17.12 following the suggested deployment deploy-1.17-and-later I noticed that Node Affinity for pv's topology.hostpath.csi/node gets set to the instance group of the pool that is assigned the hostpath pods. I can control where these pods are run by adding a selector to the plugin spec but it would be nice to utilize hostpath.csi.k8s.io accross multiple pools in my cluster.

Is there a suggested way to deploy csi-driver-host-path so it works across all node pools?

[pohly]

Is there a suggested way to deploy csi-driver-host-path so it works across all node pools?

Not at the moment. Provisioning will be problematic once there is more than one instance of the driver in the cluster. An idea how to address that is tracked in kubernetes-csi/external-provisioner#487

[nemo83]

Hi,

not sure if this is the same issue as above, but I currently have this scenario:

ubuntu@pi-one:/var/snap/microk8s/current/args$ k get csinode
NAME       DRIVERS   AGE
pi-four    0         5d17h
pi-three   0         13h
pi-one     1         5d17h

And I can see that PVC are successfully created only on pi-one. I've created this helm chart with the required config to run CSI localhostpath with snapshotter and controller, https://github.com/speedwing/csi-hostpath-helm but I'm struggling to make it to work across all my raspberry pis (I'm running on microk8s if that helps)

I followed these instructions: https://github.com/kubernetes-csi/csi-driver-host-path/blob/master/docs/deploy-1.17-and-later.md

Thanks in advance

[fejta-bot]

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-contributor-experience at kubernetes/community.
/lifecycle stale

[pohly]

/remove-lifecycle stale
/lifecycle frozen

Since v1.6.0 and PR #248, "distributed provisioning" is supported. What's missing is distributed snapshotting and resizing. In both cases the sidecar must be updated somehow - no specific plans for that at the moment.

Issues for these sidecars:

• distributed resizing external-resizer#142
• distributed snapshotting external-snapshotter#484