VM Performance Dies after a few minutes with LG running

Hello,
I am currently in the process of creating a GPU passthrough system and I’m using lookingglass to view it. I have followed the installation guide and have been able to get it running with autostart, scream over LAN, etc.

After a few minutes (Anywhere from 5-30) the performance of the guest OS tanks. The mouse will skip around, programs will not open, the start menu takes over a minute to open, etc. In one case video just cut out entirely and I was left with a black screen, but I haven’t been able to reproduce that issue so it may have been a fluke. But if I boot the guest and immediately kill the looking-glass-host process with task manager, viewing the VM by plugging the GPU into a monitor, the guest will never deteriorate.

I am a Looking Glass and KVM noob so I am at a loss for ideas about how to fix it. I’m not sure if I’ve misconfigured something or if there is a problem with the application. I can provide any files needed and if you need more information I’ll do my best to give it.

Thank you in advance for any help.

From what you describe I assume you are running on the bleeding edge release as I have also seen this behaviour but have not yet managed to track it down.

Just to be sure I redownloaded and I can say with certainty that I’m experiencing this on the official release B1. As I type this from my host OS, I’m running the VM on a second monitor and it is “working” albeit with the stuttering issues I was describing earlier. On the host OS, the looking glass client is actually showing an image from several minutes ago and not updating at all.

Since this seems to be happening after a delay, maybe it’s a possible issue with the shared memory space not being cleaned out after each frame?

Impossibe, there is nothing to ‘clear out’. It sounds more like the host application is crashing/shutting down, and/or hanging. Please describe your setup, both hardware and software.

My hardware is a Ryzen 1600X on an X370 board, 16 gigs of RAM, with a GT 730 for the host and a GTX 970 for the guest. I’m running a clean Arch install and my VM was set up by following the “PCI passthrough via OVMF” guide on the Arch wiki (the forum isn’t letting me post links).

A lot of sources I’ve checked have said you should leave one CPU core to the host, but I recently changed it to give all 12 cores to the guest in Virt-manager and so far I’ve actually not had serious problems. I’m not totally positive yet but that may have been the issue.

So it’s fixed?

If it happens again please post the host log at %TEMP%\looking-glass-host.txt.

I spoke too soon, immediately after posting it happened again.

looking-glass-host.txt (80.9 KB)

Here is my log.

Nothing obvious stands out. Can you please note exactly what steps is reproducing this issue?

After I boot the VM I just try to use it as you would a normal computer. Basic web browsing, system settings, etc. In this most recent case I watched YouTube for about half an hour before posting, after posting I opened Minecraft to check if the video quality was as it seemed and that’s when it died. I doubt that’s the direct cause though because it has happened without opening any games before.

Something important to note, the VM is actually locking up, it’s not just the looking glass client not getting any new info. I have the VM plugged directly into a second monitor and it freezes at the same times, though if it goes on long enough the client will stop updating the frames outright while the VM monitor continues in its stuttering way.