Kvdb wrong path after node restart: trying to reach Flashblade NFS Endpoint to reach VMDK px-clouddrive


portworx won’t start on my clusters after node maintenance.

We are using vsphere cloud provider + pure-secret with a flashblade mostly for license purposes.

It tries to find kvdb on the FB NFS Endpoint, this is obviously a wrong behavior, since we don’t use the NFS Endpoint for kvdb. See the vmdk path, it should never ever have a NFS path.

message: 'Failed to start Portworx: error loading node identity: Cause: ProviderInternal
Error: volume [datastore-306] fcd/1fe5017548b644ae8b59f3defd0b1dca.vmdk is attached
at an unsupported path[datastore-306] fcd/1fe5017548b644ae8b59f3defd0b1dca.vmdk.

I have no idea where to start.

Seems vsphere credentials expiration did a mess.
Support is investigating why suddenly the vmdk, which should have been attached at a vmware level, and resulting in a /dev/sdb cloud drive, is suddenly considered to be accessible from the VM, through the direct access NFS path (10.x.x.x), which is non-sense