Server Crash while using PCI passthrough

[speedgaze]

Hi

We are observing frequent crashes of Proxmox PVE when using with PCI Passthrough for NVIDIA GPUs. Below is the error we get on console

Trying to get more telemetry data to help us understand this better, but can someone throw some light on this?

 

[leesteken]

We are observing frequent crashes of Proxmox PVE when using with PCI Passthrough for NVIDIA GPUs. Below is the error we get on console
View attachment 67002
Trying to get more telemetry data to help us understand this better, but can someone throw some light on this?

Since you are breaking the virtualization barrier with PCI(e) passthrough, the VM can take down the Proxmox host (via the shared PCIe bus) when a driver or device malfunctions.
Unfortunately, I have no idea why your crash happens and I don't see any useful information on the screenshot to search the internet for. If this happens a lot, you might try (temporarily) using a different GPU (or even test and/or replace other hardware parts such a memory) to see if this help isolating the isssue.

 

[zzz09700]

Looks like a kernel panic to me.
Try grab a kernel crash dump from host or guest and see if you can find more useful info there.