Crashes

[neuroscr]

One of my KVM/QEMU die under certain Database software processing (on PMVE 1.0 & 1.1), but there's no logs of anything. No debugging or helpful crash info. I checked everything in /var/log, only a note of a nic disconnect of the VM. Also there's no option to automatically restart a crashed KVM/QEMU.

Thoughts, Comments, Suggestions?

 

[tom]

One of my KVM/QEMU die under certain Database software processing (on PMVE 1.0 & 1.1), but there's no logs of anything. No debugging or helpful crash info. I checked everything in /var/log, only a note of a nic disconnect of the VM. Also there's no option to automatically restart a crashed KVM/QEMU.

Thoughts, Comments, Suggestions?

please see the KVM project for howtos and debugging infos, also the lists are quite helpful (there is also an archive).

your setting of the KVM guest (post your /etc/qemu-server/VMID.conf)?

 

[mangoo]

One of my KVM/QEMU die under certain Database software processing (on PMVE 1.0 & 1.1), but there's no logs of anything. No debugging or helpful crash info. I checked everything in /var/log, only a note of a nic disconnect of the VM. Also there's no option to automatically restart a crashed KVM/QEMU.

Thoughts, Comments, Suggestions?

Certainly, crashes should not happen.

Automatic restart if machine crashes? Could be tricky to distinguish between a normal crash and a regular machine "stop/shutdown" (are pidfiles removed when it crashes for you?).

If you're unable to debug these crashes, you can always automatically "restart" guests with a simple script started in cron, i.e.:

ps u -C kvm | grep -q <VMID>.pid
if [ $? -ne 0 ] ; then
/usr/bin/kvm -monitor unix:/var/run/qemu-server/<VMID>.mon......
fi

 

[mfa2004]

I have been running Proxmox in production for the past month and have been generally happy with it overall.

However, in this same period I have encountered instances when a running KVM instance dies out. Unfortunately, I can't seem to find any log that might direct me to what is wrong.

Personally, I suspect that during times of high load (more likely high IO load, such as when the machine is doing a full system scan for viruses and it so happens that another VM is still doing the same thing) these things happen.

Is there something I can enable somewhere to record what's happening (a more detailed log)?

As a temporary workaround, the script above provides a nice skeleton for recovery (just start the VM that died) ... but is there someway I can call the same parameters that are passed to kvm by proxmox, so that it will always be consistent with running the VM from the Web UI?

Again great work and kudos to the entire ProxMox team ... more power to you!

 

[dietmar]

As a temporary workaround, the script above provides a nice skeleton for recovery (just start the VM that died) ... but is there someway I can call the same parameters that are passed to kvm by proxmox, so that it will always be consistent with running the VM from the Web UI?

qm showcmd <VMID>

shows you the exact command line.

 

[mfa2004]

PERFECT!

Exactly what I was looking for ... thanks!

 

[neuroscr]

Thank you, I stitch them together to make this:

#!/bin/sh
ps u -C kvm | grep -q VEID.pid
if [ $? -ne 0 ] ; then
`/usr/sbin/qm showcmd VEID`;
echo "VEID restart"|mail email@domain.com
fi

seems to work good. I had the kvm die twice within 30 minutes tonight. I was using the virtio driver, I'm switching it realtek to see if its the virtio driver.