[SOLVED] Weirdes problem ever with NFS

[Kei]

Hello guys,
I am wondering if it is possible to manually mount (in /etc/fstab) a NFS volume avoiding the GUI, and then creating a shared local directory from the GUI, that points to the NFS path.
Is this a safe procedure for shared storage, provided that I edit all the /etc/fstab of all nodes in the cluster?

This apparently ridicolous question, is due to the fact that I'm looking for a workaround for the following problem that I cannot solve:

I am using FreeNAS as storage, and the problem is that, for some reason, there seems to be a protocol miscomunication between PVE and FreeNAS, such that the FreeNAS server get hogged with CLOSE_WAIT connections, until the quewe buffer is full. The /var/log/messages shows this:

kernel: sonewconn: pcb 0xfffff800853dc620: Listen queue overflow: 193 already in queue awaiting acceptance

This problem is already acknowledged by the community and still we can't understand if the falut has to be looked for on PVE or FN (my guess is FN). Still, I need a workaround for this.

Please help!

 

[udo]

Hello guys,
I am wondering if it is possible to manually mount (in /etc/fstab) a NFS volume avoiding the GUI, and then creating a shared local directory from the GUI, that points to the NFS path.
Is this a safe procedure for shared storage, provided that I edit all the /etc/fstab of all nodes in the cluster?

Hi,
yes this should work without problems.

Udo

 

[Kei]

Thank you for the reply Udo.
I managed to make some more research and apparently solved the issue just by adding this line to every NFS block in /etc/pve/storage.cfg:

options rw,relatime,vers=4.0,rsize=131072,wsize=131072,namlen=255,hard,proto=tcp,port=0,timeo=600,retrans=2,sec=sys

Dunno if everyone of these parameters are needed, but this way it seems to be working.