VMs try to start before NFS is mounted

[Benjamin Hodgens]

I've got a problem with proxmox 4.1 (which also existed on 3.3) where VMs try to start before the NFS filesystems have mounted, specifically on a host reboot. On 3.3, I worked around the issue by increasing the NFS server threads, but that appears to do nothing for 4.4.

For instance, if I reboot the host, NFS will mount - but it takes a while. I've tried fiddling with VM startup delays and things like that, to no avail.

I am able to "instantly" mount the NFS exports manually via bash without a problem, but setting them up through proxmox appears to have them take a significantly longer time - long enough for me to have to login via ssh to check their status.

So I might investigate this further, does anyone know what is the mechanism used to initiate the mounting of NFS? Is there any way to increase verbosity of loggin?

 

[wolfgang]

Hi,
does it help to set a startup delay on the VM's?

 

[Benjamin Hodgens]

It doesn't, unless I use a "sacrificial" VM using NFS storage, which I don't care about starting, and then make it so the others all delay 60s+ from that one. It's a really ugly fix, and doesn't work consistently.

For what it matters, this issue is not tied to DNS - NFS is mounted via IP.