This message was deleted.
# longhorn-storage
a
This message was deleted.
i
Can you elaborate more on your question?
c
Again and again Longhorn stalls with "nfs server not responding". See my posting of yesterday. I would like to disable Longhorn's RWX in favor of the old nfs-provisioner and a dedicated NAS.
i
Do you have RWX volumes in your longhorn cluster?
BTW, what’s the Longhorn version you’r using?
c
Yes, there are RWX volumes, but they can be removed. Longhorn is version 1.4.0, installed via the longhorn-crd-101.2.0+up1.4.0 and longhorn-101.2.0+up1.4.0 helm charts in Rancher.
i
A little confused. If you don’t want to use the Longhorn RWX volumes. Why not just use the old nfs-provisioner instead rather than create RWX volumes?
c
Because longhorn is the default storage class for all access modes. I have only limited control over which storage classes my colleagues use. I just don't want to provide longhorn as a default for RWX anymore.
i
BTW, do you experience the
nfs: server 10.43.170.7 not responding, timed out
in previous version?
And what’s your worker node’s OS version?
Because longhorn is the default storage class for all access modes. I have only limited control over which storage classes my colleagues use. I just don’t want to provide longhorn as a default for RWX anymore.
There is no way to disable the Longhorn RWX. I think we can investigate the NFS version not responding issue and see if there is any issue in the OS layer. Do you know how to reproduce this issue and can you help provide more information of your cluster? Thank you.
c
I haven't found a reliable way to reproduce this problem, yet. By now it came up just a few times. If an NFS server pod dies for whatever reason, is it restarted or replaced?
i
Got it.
NFS server pod died issue
is caused by the crash of the internal nfs-ganesha server. We’ve fixed it in v1.4.1. The NFS server pod should started automatically, but the crash is uncertain and probably results in a dead mountpoint.
c
Good to know. I have found #5456 on github, but I am not sure if this is the correct issue.
i
Yeah, that’s it.
c
Are you sure it has been fixed for 1.4.1? It was closed 3 days ago. Milestone is 1.5.0.
i
Yeah, it is backported.