Proxmox Becomes Unreachable

[tsadigov]

I hope there is an easy approach to handle my torture. I had a proxmox setup used till a week ago.

My setup:
I use GPU passthrough so that my AI guest VM can use gpu accelaration. Because of the settings required by pci passthrough vga output from host is disabled. I have 2 ethernet ports on the motherboard.

Problem:
Suddenly my proxmox became unreachable. The web ui does not work. I can not ssh into it. I can ping it.

What I did:
I ended up reinstalling proxmox from scratch. I also switched to second ethernet port. Which looked fine. I copied some db files from the old storage to the new one so my VM s are there.

Result I got:
Initially it looks like everything is working. I can boot my existing VM s except the one requiring pci passthrough. But when I try to set up the pci passthrough somehow I manage to lock the system in a similar way that it was locked initially. It is 6 th time I am reinstalling the proxmox and I strart wondering if it has something to do with disabling nvidia drivers or some modprobe changes. It feels like I am halucinating. Wanted to ask sane ppl

 

[nunner]

Hi,

can you verify that your network controller and your GPU are in different IOMMU groups? Please post the output of this [1] script.

[1] https://wiki.archlinux.org/title/PCI_passthrough_via_OVMF#Ensuring_that_the_groups_are_valid

 

[tsadigov]

Hi,

can you verify that your network controller and your GPU are in different IOMMU groups? Please post the output of this [1] script.

[1] https://wiki.archlinux.org/title/PCI_passthrough_via_OVMF#Ensuring_that_the_groups_are_valid

Thanks for suggestion.
It seems to be better now, I added pcie_acs_override=downstream ... into grub
So I assume your point was right
And that may explain why I was able ping but not ssh into the machine. Probably I was talking to the guest machine that is automatically started.
I cant go back to failing setup to put run that script, but will run it if I will have to set up from scratch so that search engines index it also .

Again thank you @nunner

 

[nunner]

Glad to hear that. Please mark the thread as solved so that others with the same problem can find it more easily . Note that enabling that option has certain implications [1][2].

[1] https://wiki.archlinux.org/title/PC...passing_the_IOMMU_groups_(ACS_override_patch)
[2] https://vfio.blogspot.com/2014/08/iommu-groups-inside-and-out.html