Entire PVE unresponsive when working in VM

[surfninjas]

- Proxmox 9.1.5
- kernel: 6.17.9-1-pve
- VM OS: ubuntu server 24.04.3
- hw.txt(attached) contains output of lshw -short

Hi Proxmox Community!
I have noticed that sometimes when I am running once of my VMs, after a little time working in the VM, almost certainly it will freeze up and will cause my entire PVE to have a connection error. When this happens, I cannot ssh or ping the machine from my laptop (accessing the PVE web UI from my laptop), but if I connect a mouse, keyboard, and monitor directly to the server I can login as root with not issues. The error displayed on the PVE WebUI says "connection error."

Here are some of the VM configurations I have tried after searching this forum, without success:

• Switched Processors Type to host
• Enable Numa (in processors)
• Turned off Allow KSM in Memory tab
• Set BIOS to defaults and then turn on the proxmox ones
• SCSI Controller Type to LSI default
• SSD Emulation on

The VM configuration is attached in the file vm.txt. I was also able to capture a log from the machine using journalstl -xe (crash.log).

Any pointers you can give me is greatly appreciated as this has been an issue for several weeks now. It's unclear if there is a particular action that I take that will cause it freeze like this and it can happen almos at any time. I am also happy to provide more information if that would be useful.

Thank you in advance for your help!

P.S.
I'm also aware that my NAS is offline. I can access it through a different URL but the proxmox page continually tells me it's unavailable. Any pointers on how to fix this is also much appreciated.

 

[fstrankowski]

I think thats a pretty pretty old bug IMHO. Seen that couple of times already and encountered a similiar problem with an e1000 driver on one of my mini PCs aswell. I've looked up the old thread regarding this and its here.

TLDR;

Its most likely related to an old intel network card bug. Disable TSO for your network card -> put this into your /etc/network/interfaces at the appropriate section.

post-up ethtool -K nic0 tso off

if its still problematic (most likely not), change it to

post-up ethtool -K nic0 tso off gso off

TSO = TCP Segmentation Offload
GSO = Generic Segmentation Offload

Edit: If you're interested in the backstory, here is the original Kernel.org Bugzilla Bug which is still open after 13 years ;-)
Personal note: I was sick of workarounds and use a 2.5G USB 3 nic on that PC instead.