csi-gcs icon indicating copy to clipboard operation
csi-gcs copied to clipboard

Unable to attach or mount volumes

Open jailbreakerSN opened this issue 3 years ago • 11 comments

Hello,

I'm facing a issu when trying to test csi-gcs on GKE after creating the secret csi-gcs-secret

  • kubectl create secret generic csi-gcs-secret --from-literal=bucket=<BUCKET_NAME> --from-file=key=<PATH_TO_SERVICE_ACCOUNT_KEY>
  • kubectl apply -k "github.com/ofek/csi-gcs/examples/static?ref=v0.7.0" The errors I got are:
  • AttachVolume.Attach failed for volume "csi-gcs-pv" : attachdetachment timeout for volume csi-gcs
  • Unable to attach or mount volumes: unmounted volumes=[csi-gcs-pvc], unattached volumes=[csi-gcs-pvc default-token-n6w8k]: timed out waiting for the condition

jailbreakerSN avatar Apr 30 '21 19:04 jailbreakerSN

@ofek can you check please my issue? Thanks in advance!

jailbreakerSN avatar May 10 '21 15:05 jailbreakerSN

@dysosmus Do you think this might be related to https://github.com/ofek/csi-gcs/pull/72?

ofek avatar May 11 '21 16:05 ofek

@ofek I can observe similar events in our clusters and it's related to the driver starting up. In our case, those warnings are however transitory and doesn't prevent the pod from eventually start.

dysosmus avatar May 11 '21 19:05 dysosmus

Or course @dysosmus, it doesn't prevent the pod from start, but stay at containairizing step, because it seems to wait for a volume that never provided !

jailbreakerSN avatar May 11 '21 21:05 jailbreakerSN

it doesn't prevent the pod from start, but stay at containairizing step, because it seems to wait for a volume that never provided !

@ofek Then that seem different, from what I'm working on #72 :( . In my experience, the container eventually start after ±1-2 minutes of retries.

@jailbreakerSN is the csi-gcs driver pod is running properly on the node where your pod is running?

dysosmus avatar May 13 '21 08:05 dysosmus

Hello @dysosmus , Yes, I think all is running (same namespace, same node) like the commands I specified earlier. I put the Key and bucket Bâle correctly and ran out with kubectl

jailbreakerSN avatar May 13 '21 12:05 jailbreakerSN

@maennchen Do you have any ideas?

ofek avatar May 17 '21 23:05 ofek

@ofek @dysosmus , might it be related to the status of my cluster, which is a private cluster?

jailbreakerSN avatar May 17 '21 23:05 jailbreakerSN

@jailbreakerSN could be, do your subnet where your GKE cluster nodes are running have Google Private Access enabled? You can also try to spin up a GCP Cloud NAT on the subnet, but afaik that shouldn't be needed if Private Access is enabled.

dysosmus avatar May 18 '21 09:05 dysosmus

@jailbreakerSN Can you capture the log of the csi-gcs DaemonSet pods while you provision a new PVC and start a pod using it?

maennchen avatar May 18 '21 09:05 maennchen

@jailbreakerSN , Is this issue still present? even I want to try this in my GKE services

elamcheliyan avatar Jan 18 '22 11:01 elamcheliyan