Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

pvcviewer-controller-manager creates pvcviewer pod on different node from notebook server pod #60

Open
muratyarali opened this issue Jul 1, 2024 · 1 comment

Comments

@muratyarali
Copy link

I'd like to ask you something about the pvcviewer-controller-manager.

1- I want to see the volume content attached to a notebook by clicking the "Open PVCViewer" icon on Volumes page.
Screenshot_13

2- But after that, it is waiting forever to open the volume content on Kubeflow UI and says "The PVCViewer is currently starting"
Screenshot_14

3- When i checked the details from the CLI, The pods that belong to the notebook and pvcviewer are created and they are on different nodes.

$ k get po -A -o wide | grep e2e-mlops
kubeflow-user-example-com              e2e-mlops-0                                      2/2     Running   0  43h     10.2.11.25   node-3
kubeflow-user-example-com              pvcviewer-e2e-mlops-workspace-7fbcb66bc7-sd2xd   0/2     Init:0/1  0  2m33s   <none>       node-2

4- When i describe pvcviewer-e2e-mlops-workspace-7fbcb66bc7-sd2xd pod, i can see that there is "Multi-Attach error"

$ k describe po -n kubeflow-user-example-com pvcviewer-e2e-mlops-workspace-7fbcb66bc7-sd2xd

Events:
  Type     Reason              Age                 From                     Message
  ----     ------              ----                ----                     -------
  Warning  FailedAttachVolume  8m4s                attachdetach-controller  Multi-Attach error for volume "pvc-d3e10e54-33ee-4832-97b4-ee7895ddb471" Volume is already used by pod(s) e2e-mlops-0
  Warning  FailedMount         89s (x3 over 6m1s)  kubelet                  Unable to attach or mount volumes: unmounted volumes=[viewer-volume], unattached volumes=[viewer-volume], failed to process volumes=[]: timed out waiting for the condition

5- When i check the related persistent volume

$ k get pv pvc-d3e10e54-33ee-4832-97b4-ee7895ddb471 -o wide
NAME                                      CAPACITY  ACCESS MODES  RECLAIM POLICY  STATUS  CLAIM                                          STORAGECLASS
pvc-d3e10e54-33ee-4832-97b4-ee7895ddb471  5Gi       RWO           Delete          Bound   kubeflow-user-example-com/e2e-mlops-workspace  csi-rbd-sc

5- The problem here is;
csi-rbd-sc storage class can only create a volume with RWO access mode so second pod (pvcviewer-e2e-mlops-workspace-7fbcb66bc7-sd2xd) can not connect the volume and gets Multi-Attach error.

The only possible solution here is I must configure pvcviewer-controller-manager to create pvcviewer pod in the exact same node with notebook server pod (e2e-mlops-0). I tested it with manuel configuration and it works.

@thesuperzapper can you track a fix?

@andreyvelich
Copy link
Member

/transfer dashboard

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
Status: To Do
Development

No branches or pull requests

2 participants