[SOLVED] Windows VM reboot or shutdown causing 100% cpu on host

[silencery]

Hi proxmox community,

I've encountered a strange situation after updating to 6.8.12-1-pve and I'm hoping someone can help.
I recently tried to update a windows VM. Once Windows Update finished, I discovered the entire host crashes.

Digging a little deeper, i noticed that when issuing any shutdown command on this client, the proxmox host suddenly shoots to 100% CPU.
The only way to recover is a hard reboot of the host since even the shell becomes non-responsive.

This behavior seems to have only begun after a recent proxmox update (I went from 6.5 kernel to 6.8). I have not found anything in the logs which may point to the cause, but honestly, I'm not even sure which files to look at.

Where do I start to troubleshoot this?
Any pointers in the right direction would be appreciated.

Thanks!

 

[_gabriel]

This behavior seems to have only begun after a recent proxmox update (I went from 6.5 kernel to 6.8).

confirm with rebooting with 6.5.
is VM stored on ZFS ?

 

[silencery]

Oh, do you mean this exists on 6.5 too? I didn't notice before.

Yes, that's correct. VM is on ZFS.

 

[_gabriel]

do you mean this exists on 6.5 too?

no, you need to reboot into 6.5 kernel to confirm if it's a 6.8 issue.
https://pve.proxmox.com/wiki/Host_Bootloader#sysboot_kernel_pin

 

[silencery]

awesome, thanks! will try that

 

[silencery]

Thanks to your advice, I was able to confirm it is happening with 6.5 kernel too, so I can rule out 6.8 kernel.

Looks like I need to check ZFS now? Any tips for what to look for would be greatly appreciated.

 

[_gabriel]

There is an issue about virtio scsi driver, downgrade to version 0.1.208 to fix it

https://forum.proxmox.com/threads/r...device-system-unresponsive.139160/post-691337

Then, as always, ZFS is angry, what your exact model disks ?

Don't forget to edit your title to solved to help others.

 

[silencery]

Thanks so much!

After trying out your tips and hours of troubleshooting, i was finally able to narrow it down to a problem with a Google Coral PCI passthrough device (I've posted in the coral thread to see if anyone else can replicate this). Disabling the device allows the windows VM to start normally now.

It took much longer to troubleshoot since cold start is fine, and the problem only showed when rebooting the VM.
I had to resort to blindly changing configs since I didn't even know which logs to check. For future reference, is there a checklist of logs to check when something like this happens?