[PVE-User] How to troubleshoot SCSI disk issue under load

Chris Murray chrismurray84 at gmail.com
Mon Jul 29 22:10:18 CEST 2013

Hi all,

I'm having some trouble with a V2V'd Windows Server 2008 VM which was
working fine on vSphere 5.1. Apologies, I'm a bit of a Proxmox novice,
and there do seem to be reports on the 'net of complaints similar to
mine, although I'm not sure how to troubleshoot my own issue and would
appreciate some advice.

I booted using a systemrestore cd, and DD'd a number of disks across to
Proxmox. On the destination VM, there's one IDE drive for booting from,
and a number of SCSI disks with writethrough cache. I admit, I did
forget to uninstall VMtools prior to this process, but have since
removed them.

Under load on the SCSI disks, the virtual machine will switch off (not
BSOD or restart). I can reproduce this easily with a "chkdsk x: /r" for
the SCSI disk, but not with a "chkdsk c: /r" (and subsequent reboot) for
the IDE disk. That would point me in the direction of a problem with the
SCSI driver/virtual controller?

Thinking that it was something to do with drivers, I removed a disk,
re-added it as a Virtio device, installed drivers from
virtio-win-0.1-52.iso and the system suffered a BSOD. It would BSOD
before even booting into safe mode also. It even did this with all of
the SCSI & Virtio disks disconnected. Only booting with Last Known Good
configuration got me back into windows, and back to a position where a
"chkdsk x: /r" would turn off the VM, rather than restart it or BSOD.

I removed this disk and added as IDE again. chkdsk was now fine.
Removed and added again as SCSI. Machine powers off under load.

In case it's anything to do with not being on the most recent build, I
upgraded one host with "apt-get update" and "apt-get dist-upgrade",
migrated the machine to it, however this does still happen.

Just in case the file format was having any sort of effect, I migrated
from raw to qcow2. Unsurprisingly, this hasn't made any difference.

Any ideas where to turn next? I'm intrigued by the machine powering down
rather than suffering a BSOD.

host pm00 has kernel Linux 2.6.32-20-pve
host pm01 has kernel Linus 2.6.32-22-pve 

Many thanks,

More information about the pve-user mailing list