PVE became very slow when the NFS server used for backup crashed

[f4242]

Hello,

My PVE servers were very slow this week-end. The root cause is the NFS server I use for backup crapped out (very sluggish and then it crashed completely). The NFS server is only used for backup, the VM data is on local disks.

I was unable to stop the backup tasks. The stop button on the web UI just did nothing.

Some VM were affected, but not all. The VMs became very slow. I couldn't even SSH some VMs (the ssh client timed out).

Once my NFS server restarted and become available again, everything come back to the normal. The VM that were not accessible by SSH started to answer again, without the need of rebooting it.

I rebooted one PVE host on the week-end to restore service on it. This server worked fine after the reboot even if the NFS server was still down. I think the problem could be caused by the crash of the NFS server when pve-backup was running.

I wonder if there is something I can do to avoid that issue? A backup job should not make the VMs unresponsive like that.

Thanks.

pve-manager/5.1-46/ae8241d4 (running kernel: 4.13.16-1-pve)

 

[dietmar]

I would simply use a reliable/fast backup storage ... It may also help if you use the 'soft' mount NFS option.

 

[fireon]

I wonder if there is something I can do to avoid that issue? A backup job should not make the VMs unresponsive like that.

Or you change to CIFS.