You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Use this startup script to ensure Nvidia is passed through correctly nvidia-persistenced; /mnt/{your_data_set}/jailmaker/jlmkr.py startup
I installed dockage which is how I'm managing the docker compose files
The issue I was running into was installing the Nvidia drivers, because this version of Truenas comes with 545.23.08 which is not avaliable on Nvidias website
I can see in the 60-configure_gpu_driver.sh it tries to fall back to using Github, but it doesn't work for some reason, and also if it did work it would pull the wrong version of the driver NVIDIA-Linux-aarch64- instead of x86
Once I had this file, I put it in a dataset and mounted it into /home/default/Downloads and renamed it to NVIDIA_545.23.08.run which is the format the script looks for before it tries to download it.
After that it installed the driver from that file and everything worked correctly!
I believe the issue with the 60-configure_gpu_driver.sh script is that it creates the NVIDIA.run file in /tmp even tho it didn't download the driver. This caused my container to exit when it tried installing with exit code 0 which was really hard to debug. I figured this out by not passing through the Nvidia drivers which let me run the container with integrated graphics, and then I ran that script manually and looked at the output in /tmp, the file size was 0.
Anyways, I hope this helps someone get the install working on Truenas scale, this is super cool and thinks for all the hard work!
The text was updated successfully, but these errors were encountered:
jayfry1077
changed the title
Tuenas Scale - Getting Nvidia drives to install on Dragonfish-24.04.2
Tuenas Scale - Getting Nvidia drivers 545.23.08 to install on Dragonfish-24.04.2
Aug 3, 2024
Description
Hello, I wanted to share what I had to do to get this working on Truenas scale Dragonfish-24.04.2
First I followed this tutorial to get my jail running with docker and Nvidia GPU pass through
https://www.youtube.com/watch?v=S0nTRvAHAP8
Use this startup script to ensure Nvidia is passed through correctly
nvidia-persistenced; /mnt/{your_data_set}/jailmaker/jlmkr.py startup
I installed dockage which is how I'm managing the docker compose files
The issue I was running into was installing the Nvidia drivers, because this version of Truenas comes with 545.23.08 which is not avaliable on Nvidias website
I can see in the 60-configure_gpu_driver.sh it tries to fall back to using Github, but it doesn't work for some reason, and also if it did work it would pull the wrong version of the driver NVIDIA-Linux-aarch64- instead of x86
I was able to download the driver I needed from github tho
https://github.com/flathub/org.freedesktop.Platform.GL.nvidia/releases/download/cuda/NVIDIA-Linux-x86_64-545.23.08.run
Once I had this file, I put it in a dataset and mounted it into /home/default/Downloads and renamed it to NVIDIA_545.23.08.run which is the format the script looks for before it tries to download it.
After that it installed the driver from that file and everything worked correctly!
I believe the issue with the 60-configure_gpu_driver.sh script is that it creates the NVIDIA.run file in /tmp even tho it didn't download the driver. This caused my container to exit when it tried installing with exit code 0 which was really hard to debug. I figured this out by not passing through the Nvidia drivers which let me run the container with integrated graphics, and then I ran that script manually and looked at the output in /tmp, the file size was 0.
Anyways, I hope this helps someone get the install working on Truenas scale, this is super cool and thinks for all the hard work!
The text was updated successfully, but these errors were encountered: