Looking Glass - Triage

Okay, gonna wait for Beta 1 before doing the video. Tend to want to do release code demonstrations vs cloning from Git if you don’t mind.

What’s the benefit? I don’t see a reason to switch if there isn’t an issue or feature we don’t have.

1 Like

That’s something I’m hoping someone can answer. Unknown Worlds (Subnautica Developer) moved to Favro because it allows a singular page to show multiple collections.

Ah, okay. I wasn’t sure if it was a suggestion to follow suit or just a comment.

(I’m catching up on the forum in rapid succession, sorry!)

Not sure if I’m having a Looking Glass problem, or a SPICE problem, but I thought I’d ask here because I’m not really sure where to ask about SPICE problems. I’ve got LG all set up and everything is working except for mouse input. The mouse is stuck at the top of the screen at all times. I know spice is detecting it because right click works, but I can’t move the mouse or anything. I did install the registry hack to disable mouse acceleration, but the problem stayed the same. Keyboard input works as smooth as butter, though, as does passing the mouse directly through to the machine.

Client and host are both running at the same resolution, and I’ve tried it in fullscreen as well as windowed.

Any ideas?

Remove the tablet device from the VM

2 Likes

Fixed it! Thanks!

Hello, sort of new to linux and I’ve followed the quick start guide for looking glass but I wasn’t able to do chown user:kvm /dev/shm/looking-glass, kept saying invalid user user:kvm. So I tried chmod 777 /dev/shm/looking-glass, but that didn’t work either. Is it possible to move the looking glass file to say, the documents folder to get it to work? Also after this is all done, is there anything else that I should do to get the best performance possible? Like tuning the VM or something?

Did you replace ‘user’ with your actual username?

Edit: Also make sure that your user is part of the KVM group

sudo usermod -a -G kvm username

Thanks and no, I didn’t, silly me. But unfortunately it still says permission denied. Also I’ve been searching and found this.

Problem:
The VM won’t start because QEMU/libvirt can’t stat /dev/shm/looking-glass or otherwise can’t read it

Solution:
Either /dev/shm/looking-glass doesn’t exist (has to be recreated every boot!), file permissions are incorrect for QEMU/libvirt to Read/Write to it, AppArmor is enabled and /etc/apparmor.d/abstractions/libvirt-qemu is missing the /dev/shm/looking-glass rw, rule, or the AppArmor service hasn’t been restarted after adding that rule.

How do I add that rule, how do I restart it, and how do I make it so it remakes the file everytime?

1 Like

Assuming it’s the apparmor rule that’s missing: open up the /etc/apparmor.d/abstractions/libvirt-qemu file in a text editor and add to it

/dev/shm/looking-glass rw

Then reload aa

sudo systemctl reload apparmor

2 Likes

Yep, that did it, thanks! Now how do I make it so that the looking-glass file gets remade after every restart?

You could add the scripts to your /etc/rc.local file, which gets executed at boot.
If the rc.local file doesn’t exist, create it and add this:

https://pastebin.com/5MU1KPKh

(replace ‘username’ with your actual one)
And make it executable.

sudo chmod +x /etc/rc.local

‘touch’ creates an empty file, and you wanted to chown the file afterwards.
In the above example ‘install’ creates also an empty file, but with file permissions set accordingly in a neat oneliner.

EDIT: Edited for clarity

2 Likes

Didn’t find information about GPU resource sharing to the number of virtual machines.

I am searching technology (and developer) for split GPU and running much possible sandboxed (VM) 3d custom applications in parallel.

It is a possible with Looking Glass (PCI passthrough)?

What you are asking for is SR-IOV, it at current is a feature only availble on professional cards like the FirePro and NVidia Tesla/GRID series. Looking Glass will work with it but it is not what the majority of people here are doing.

Thank you for answer.
My small startup company is need to stream 3D app to browser for a lot of concurrent visitors (e.g. 100 in same time), and we seek how to split Google Cloud or EC2 GPU instances (Tesla and GRID available). We decide to develop 3D apps SaaS hosting, but currently stuck on questions how much is that would be cost and if technologies is already exist to make cheap hosting for lightweight 3d apps.

No worries. It sounds like you’re doing something similar to Parsec

We also support Internet Explorer, mpeg1 and websockets.
Not games, but for learning.
Parsec use own client app, Dixper and Rainway use Chrome with webrtc, also for games and not a way to split GPU resources to pieces for lightweight apps.

Sorry, to be clear I was referring to the hosting hardware setup that they are using would be very similar to what you require.

Thank you for clarifying.
Than i’m try Looking Glass with SR-IOV on Tesla based GPU cloud instance.