Help recovering a crashing server

[ghamauricio]

Hello guys,

Newbie here. So I built this dual-seat workstation, and the VMs are set to auto start.
I passed-through some devices to a VM, and it crashes the whole server.

I tried to boot an Ubuntu live and mount the lvm, but there's no /qemu-server folder in /etc, so I can't edit the VM's config file in order to remove the devices and suspend this crashing loop.

I know, pretty dumb on my part.

Any suggestions?


EDIT: SOLUTION

Ok so I went back to Ubuntu Live, mounted the LVM, edited the /etc/hosts and /etc/hostname to change the node's host name. Rebooted. SSH'ed into it, edited the vm.conf removing the problematic PCI devices, renamed the host name back, rebooted again and voilà.

 

[nhlax2_maduna]

have you tried remaming the primary lvm / vmdk disk of the offending VM from the ubuntu, and the rebooting your proxmox just to see if boots?
fyi: lvm renaming may further damage your setup, do it at own risk.

 

[ghamauricio]

I forgot to mention, before setting auto boot and decommissioning my previous workstation, I tried several times to pass-through separately the USB controller (crash), and the on board sound card (totally laggy on the other VM, but reversible).

Even though they’re in different IOMMU groups, but on the same address (12.00.01 to 12.00.04 with some encryption controller and maybe a PCIe dummy controller), after reading somewhere that passing all the functions might be better, it didn’t strike me that auto boot could make this irreversible.

Also, the VMs receive their own NVMe (which is a pain to get to in this motherboard). I’ll try to remove a VGA or something simpler in the hope of the addresses changing and Próxmox throwing a critical error and not being able to start the VMs.

but if there’s a simpler solution, I’d like to try first.

 

[ghamauricio]

Ok so I went back to Ubuntu Live, mounted the LVM, edited the /etc/hosts and /etc/hostname to change the node's host name. Rebooted. SSH'ed into it, edited the vm.conf removing the problematic PCI devices, renamed the host name back and voilà.

Actually I tried to rename the host along the way and broke everything, having to fresh install Proxmox, but anyway, this would have worked and I'm reporting here in case someone ever needs it.

Gonna edit the OP to reflect solution.