Closed Ramaddan closed 2 weeks ago
I wonder if it has anything to do with this:
https://github.com/NVIDIA/nvidia-docker?tab=readme-ov-file
DEPRECATION NOTICE
This project has been superseded by the NVIDIA Container Toolkit.
Update (Feb 5, 2024): Tried to install the above but it did not fix my specific problem. Still stuck at the nvidia-caps
What is that device anyway? Is it for capture?
Since you now have the nvidia driver toolkit; could you try running
sudo nvidia-container-cli --load-kmods info
and see if the devices "magically appear"?
You can also try running Wolf without those extra devices but I fear that it'll probably fail. Btw, which distro are you using?
Also, beware of https://github.com/games-on-whales/wolf/issues/60 with that driver version. I'm sorry for all the troubles, but Nvidia is just plainly hostile on Linux..
Hi. Thanks for the help. No problem, used to nVidia.
Here is what I get
NVRM version: 545.29.06 CUDA version: 12.3
Device Index: 0 Device Minor: 0 Model: Quadro RTX 5000 Brand: QuadroRTX GPU UUID: GPU-(some number) Bus Location: 00000000:01:00.0 Architecture: 7.5
Not sure if that was it, but I have the devices now :-)
Both show now: nvidia-cap1 nvidia-cap2
But I think they really showed after a restart.
There was also something about nvidia container toolkit version prior to 1.12 having issues: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#installing-with-yum-or-dnf
Thanks
Will try to continue now
Getting stuck here now:
22:03:16.577852995 INFO | Gstreamer version: 1.22.0-0 22:03:16.578928224 INFO | Reading config file from: /wolf/cfg/config.toml 22:03:16.801477822 INFO | Selected H264 encoder: nvcodec 22:03:16.801576288 INFO | Selected HEVC encoder: nvcodec 22:03:16.803287280 INFO | RTSP server started on port: 48010 22:03:16.803337562 INFO | Control server started on port: 47999 22:03:16.803286092 INFO | HTTP server listening on port: 47989 22:03:16.803451998 WARN | [PULSE] Unable to connect, Access denied 22:03:16.803492654 INFO | Starting PulseAudio docker container 22:03:16.804202617 WARN | [DOCKER] Container WolfPulseAudio already present, removing first 22:03:16.805050646 INFO | HTTPS server listening on port: 47984 22:03:18.250209710 WARN | [PULSE] Unable to connect, Access denied
This was my docket CLI command:
docker run --name wolf --network=host -e XDG_RUNTIME_DIR=/tmp/sockets -v /tmp/sockets:/tmp/sockets:rw -e NVIDIA_DRIVER_VOLUME_NAME=nvidia-driver-vol -v nvidia-driver-vol:/usr/nvidia:rw -e HOST_APPS_STATE_FOLDER=/etc/wolf -v /etc/wolf/wolf:/wolf/cfg -v /var/run/docker.sock:/var/run/docker.sock:rw --device-cgroup-rule "c 13:* rmw" --device /dev/nvidia-uvm --device /dev/nvidia-uvm-tools --device /dev/dri/ --device /dev/nvidia-caps/nvidia-cap1 --device /dev/nvidia-caps/nvidia-cap2 --device /dev/nvidiactl --device /dev/nvidia0 --device /dev/nvidia-modeset --device /dev/uinput -v /dev/shm:/dev/shm:rw -v /dev/input:/dev/input:rw -v /run/udev:/run/udev:rw ghcr.io/games-on-whales/wolf:stable
I did not change anything from what was given on the Wolf website
It's not stuck, just waiting for a Moonlight client to connect!
As for the Pulse warning, it's fixed in the upcoming release; as a quick workaround, you can manually clean the /tmp/sockets
folder before launching Wolf.
Ok, Still stuck.
Moonlight freezes
First, I could not run as daemon, so that I can see the pin link to use. Is there a way to see it without running it in terminal?
Second, moonlight worked and gave me options of running some apps, then it froze with retroarch
Here are the terminal messages:
[2024-02-07 21:10:32] [ /etc/cont-init.d/10-setup_user.sh: executing... ]
[2024-02-07 21:10:32] Configure default user
[2024-02-07 21:10:32] Container running as root. Nothing to do.
[2024-02-07 21:10:32] DONE
[2024-02-07 21:10:32]
[2024-02-07 21:10:32] [ /etc/cont-init.d/15-setup_devices.sh: executing... ]
[2024-02-07 21:10:32] Configure devices
[2024-02-07 21:10:32] Exec device groups
[2024-02-07 21:10:33] Adding user 'root' to groups: gow-gid-107,root
[2024-02-07 21:10:33] DONE
[2024-02-07 21:10:33]
[2024-02-07 21:10:33] [ /etc/cont-init.d/30-nvidia.sh: executing... ]
[2024-02-07 21:10:33] Nvidia driver detected
[2024-02-07 21:10:33] [nvidia] Add Vulkan ICD
[2024-02-07 21:10:33] [nvidia] Add EGL external platform
[2024-02-07 21:10:33] [nvidia] Add egl-vendor
[2024-02-07 21:10:33] [nvidia] Add gbm backend
[2024-02-07 21:10:33]
[2024-02-07 21:10:33]
[2024-02-07 21:10:33] [ /etc/cont-init.d/init-gamescope.sh: executing... ]
[2024-02-07 21:10:33] Setting up Gamescope
[2024-02-07 21:10:33] Launching the container's startup script as user 'root'
0:00:00.030193650 221 0x55b836e3d4e0 WARN vadisplay gstvadisplay.c:316:gst_va_display_initialize:
1) I tried to stream to an android phone instead.
2) It somewhat works, at least firefox, but cannot seem to bring up any keyboard layout.
And retroarch opens, but the graphics are too bad to be able to use
So it seems the problem is still there
3) The other thing I noticed is that it controls the mouse cursor on the host machine, is that normal?
Shouldn't it be isolated from being able to do anything but stream from the host within the container, multiple instances?
4) I also noticed I have to keep executing the command after every reboot or shutdown
sudo nvidia-container-cli --load-kmods info
so that it finds the /dev/nvidia-cap1 and /dev/nvidia-cap2 devices again
Is there a way to make this more permanent?
Thanks
Hi,
I was trying to deploy Wolf, and all seems to be fine, until I reach the docker cli or compose instructions.
I get the following error and cannot continue:
docker: Error response from daemon: error gathering device information while adding custom device "/dev/nvidia-caps/nvidia-cap1": no such file or directory. ERRO[0000] error waiting for container: context canceled
I have an RTX nVidia Quadro GPU with driver version: 545.29.06
And the devices nvidia-caps do not exist on my system.
Running on Ubuntu 22.04 Kernel: Linux 5.15.0-92-lowlatency x86_64
Thanks