Missing backup NFS share stalls proxmox

[bohansen]

Hi,

We had a breakdown of a server last friday due to a missing NFS share for backup. It's a rather big logfile, but it seems the lost connection to an NFS server has something to do with it.
I noticed this is similar to this thread: http://forum.proxmox.com/threads/40...g-quot?highlight=not+responding,+still+trying. Is there a supported way for cancelling backup if an NFS share is suddenly unavailable?

The other issue in this case is that the backup NFS is (for the time being) a local openVZ machine with an NFS share. So actually I'm mounting a NFS on the host which is shared from a local openvz vm using unfs3. Could this be the root cause for the trouble? Any body else running a similar setup? As the log file show a lot more went wrong than just a missing NFS share. Previously the backup vm was placed on another proxmox host. Part of the log file is attached.

pve-manager: 1.7-10 (pve-manager/1.7/5323)
running kernel: 2.6.18-4-pve
proxmox-ve-2.6.18: 1.7-10
pve-kernel-2.6.32-3-pve: 2.6.32-14
pve-kernel-2.6.18-4-pve: 2.6.18-10
qemu-server: 1.1-25
pve-firmware: 1.0-9
libpve-storage-perl: 1.0-16
vncterm: 0.9-2
vzctl: 3.0.24-1pve4
vzdump: 1.2-9
vzprocps: 2.0.11-1dso2
vzquota: 3.0.11-1
pve-qemu-kvm-2.6.18: 0.9.1-10

Thanks in advance,
Bo

 

[whinpo]

just had the same problem today...
trying to restart my brand new VM hosting the NFS backup..

the GUI was frozen, no way to do qm list, or list /mnt/pve...
the VM were running without issue btw..

the only way i found was to modify the storage.cfg on one node where nothing was running, rebooted the node
scp the conf file from the cluster master and started the NFS server

=> the cluster was ok after that.

that means, I had a "not so good idea" to put the NFS Backup in a VM....
too bad

 

[dswartz]

I can confirm the same issue. I was not able to find anything useful in google, except vague hints that nfs writes from a host node to a VM were not throttled and might over-consume kernel memory. Not sure if I believe that. Whatever went wrong is on the nfs server, since I found that the VM was perfectly responsive, but doing anything at all that involved the file/directory would hang the process on the VM in 'D' state. In my case, on the server, I was trying to delete a several GB backup file on the nfs share. Doing so hung. Do an 'ls -l' on the nfs server VM in the directory holding that backup file also hung in 'D' state. I had to reboot everything to clear this up. Ouch...