GUI reports occasionally failed nodes

[svendsen]

Hi,

I have some weird behaviours with PVE 3.3-5.
Occasionally, GUI reports failed/red nodes in our PVE cluster. After some minutes all nodes return to green state again.
This behaviour typically occours while having connection issue with NFS (only used for backup, templates, ISOs).
I see following in message log:
kernel: ct0 nfs: server xxxxxxxxxxxx not responding, still trying

NFS server is on another location which is why it sometimes can lose connection, but should this have any influence on the overall cluster health?

I use these NFS parameters on my NFS server:
rw,async,no_subtree_check,no_root_squash

Any ideas?

 

[dietmar]

Stale nfs can delay system call. In this case, pvestatd takes too much time, so GUI reports nodes as 'red'.
This is a known problem, but so far we have no fix for that.

 

[svendsen]

Ok, that's what I thought.
I thought that async would fix these issues though...

Any ideas for alternative to NFS or TCP vs UDP connection?