Having trouble with the beegfs CSI driver in v 8.1.0

[Jack Benson]

I recently pulled the new master from GitHub as well as the image from a conversation between a Joe and a Curt, and attempted to get it working with k3s 1.33.5, but keep getting this error when i attempt to provision anything at all. Ive gotten one error free provision but cant replicate it. 

Unable to set connClientPort (BeeGFS 8) or c onnClientPortUDP (BeeGFS 7): connClientPortUDP CSI Driver

I'm not sure where I'm going wrong and logs have been silent. 

Jack

[Curt Dodds]

I haven't had a chance to try it yet. Hopefully over the next week and I'll post the results.

Curt 

--
You received this message because you are subscribed to the Google Groups "beegfs-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fhgfs-user+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/fhgfs-user/706a6112-a690-46e3-aa5f-f35be390954fn%40googlegroups.com.

[Joe McCormick]

Hi Jack,

The error `Unable to set connClientPort (BeeGFS 8) or connClientPortUDP (BeeGFS 7):` should only happen if the template client configuration file (typically at `/etc/beegfs/beegfs-client.conf` on the Kubernetes node) doesn't contain either `connClientPort` or `connClientPortUDP` (or its commented out). When you install the BeeGFS client package this file should be installed automatically and will contain that parameter.

A few questions:

(1) What does `kubectl logs csi-beegfs-controller-0` and `kubectl logs csi-beegfs-node-<HASH>` (if the pod ever gets to the point it is assigned to a node) show? If that parameter was missing from the client file I would expect to see an error like:

```
E1029 14:55:40.713368       1 server.go:191] "msg"="GRPC error" "error"="rpc error: code = Internal desc = unable to set connClientPort (BeeGFS 8) or connClientPortUDP (BeeGFS 7): connClientPortUDP not in template beegfs-client.conf file" "fullError"="unable to set connClientPort (BeeGFS 8) or connClientPortUDP (BeeGFS 7): connClientPortUDP not in template beegfs-client.conf file\nrpc error: code = Internal desc = unable to set connClientPort (BeeGFS 8) or connClientPortUDP (BeeGFS 7): connClientPortUDP not in template beegfs-client.conf file" "method"="/csi.v1.Controller/CreateVolume" "reqID"="0010" "request"="{\"capacity_range\":{\"required_bytes\":107374182400},\"name\":\"pvc-22b0cb1e\",\"volume_capabilities\":[{\"AccessType\":{\"Mount\":{}},\"access_mode\":{\"mode\":5}}]}"
```

(2) Was `connClientPortUDP CSI Driver` what was printed after the colon in the error? I would only ever expect the full error to look like `Unable to set connClientPort (BeeGFS 8) or connClientPortUDP (BeeGFS 7): connClientPortUDP not in template beegfs-client.conf file`.

(3) Are you using the dynamic (i.e., using a storage class) or static provisioning workflow?

(4) Do you have the beegfs-client-dkms (or beegfs-client) package installed on all Kubernetes nodes? Can you double check the /etc/beegfs/beegfs-client.conf file exists and contains either `connClientPort` or `connClientPortUDP` (on all nodes)?

(5) If you run `kubectl describe pod csi-beegfs-controller-0` and `kubectl describe pod csi-beegfs-node-<HASH>` can you confirm the image and image ID for the `beegfs` container?

Thanks,

~Joe