[Bug]: Container Crashing during Downloading nvidia driver

[Cribz679]

Describe the Bug

Any help is appreciated. My container seems to be crashing/restarting during downloading nvidia drivers. Any help is appreciated.

`Build: [2024-05-25 02:40:32] [master] [47f6f7a] [debian]

[ /etc/cont-init.d/10-setup_user.sh: executing... ]
**** Configure default user ****

• Setting default user uid=568(default) gid=568(default)
• Adding default user to any additional required device groups
• Adding user 'default' to group: 'video'
• Adding user 'default' to group: 'audio'
• Adding user 'default' to group: 'input'
• Adding user 'default' to group: 'pulse'
• Adding user 'default' to group: 'render' for device: /dev/input/event0
• Adding user 'default' to group: 'video' for device: /dev/dri/card0
• Adding user 'default' to group: 'avahi' for device: /dev/dri/renderD128
• Setting umask to 0022
• Create the user XDG_RUNTIME_DIR path '/tmp/.X11-unix/run'
• Setting ownership of all log files in '/home/default/.cache/log'
• Setting root password
• Setting user password
  DONE

[ /etc/cont-init.d/11-setup_sysctl_values.sh: executing... ]
**** Configure some system kernel parameters ****

• The vm.max_map_count is already greater than '524288'
  DONE

[ /etc/cont-init.d/30-configure_dbus.sh: executing... ]
**** Configure container dbus ****

• Container configured to run its own dbus
  DONE

[ /etc/cont-init.d/30-configure_udev.sh: executing... ]
**** Configure udevd ****

• Disable udevd - /sys is mounted RO
• Enable dumb-udev service
• Ensure the default user has permission to r/w on input devices
  DONE

[ /etc/cont-init.d/40-setup_locale.sh: executing... ]
**** Configure local ****

• Locales already set correctly to en_US.UTF-8 UTF-8
  DONE

[ /etc/cont-init.d/50-configure_pulseaudio.sh: executing... ]
**** Configure pulseaudio ****

• Enable pulseaudio service.
• Configure pulseaudio to pipe audio to a socket
  DONE

[ /etc/cont-init.d/60-configure_gpu_driver.sh: executing... ]
**** Found Intel device '12th Gen Intel(R) Core(TM) i5-12500T 12th Gen Intel(R) Core(TM) i5-12500T To Be Filled By O.E.M. CPU @ 1.9GHz' ****

• Mesa has already been installed into this container
  **** No AMD device found ****
  **** Found NVIDIA device 'NVIDIA GeForce RTX 4070' ****
• Downloading driver v545.23.08`

Steps to Reproduce

No response

Expected Behavior

No response

Screenshots

No response

Relevant Settings

No response

Version

Build: [2024-05-25 02:40:32] [master] [47f6f7a] [debian]

Platform

TrueNAS Scale via TrueCharts
Distribution: Debian GNU/Linux - 12 (bookworm)
Linux Kernel: 6.6.20-production+truenas unknown unknown GNU/Linux
GPU Drivers: | NVIDIA-SMI 545.23.08 Driver Version: 545.23.08 CUDA Version: 12.3 |

Relevant log output

`Build: [2024-05-25 02:40:32] [master] [47f6f7a176ee0c6f6c870c29397cc1a8d6d57839] [debian]

[ /etc/cont-init.d/10-setup_user.sh: executing... ]
**** Configure default user ****
  - Setting default user uid=568(default) gid=568(default)
  - Adding default user to any additional required device groups
  - Adding user 'default' to group: 'video'
  - Adding user 'default' to group: 'audio'
  - Adding user 'default' to group: 'input'
  - Adding user 'default' to group: 'pulse'
  - Adding user 'default' to group: 'render' for device: /dev/input/event0
  - Adding user 'default' to group: 'video' for device: /dev/dri/card0
  - Adding user 'default' to group: 'avahi' for device: /dev/dri/renderD128
  - Setting umask to 0022
  - Create the user XDG_RUNTIME_DIR path '/tmp/.X11-unix/run'
  - Setting ownership of all log files in '/home/default/.cache/log'
  - Setting root password
  - Setting user password
DONE

[ /etc/cont-init.d/11-setup_sysctl_values.sh: executing... ]
**** Configure some system kernel parameters ****
  - The vm.max_map_count is already greater than '524288'
DONE

[ /etc/cont-init.d/30-configure_dbus.sh: executing... ]
**** Configure container dbus ****
  - Container configured to run its own dbus
DONE

[ /etc/cont-init.d/30-configure_udev.sh: executing... ]
**** Configure udevd ****
  - Disable udevd - /sys is mounted RO
  - Enable dumb-udev service
  - Ensure the default user has permission to r/w on input devices
DONE

[ /etc/cont-init.d/40-setup_locale.sh: executing... ]
**** Configure local ****
  - Locales already set correctly to en_US.UTF-8 UTF-8
DONE

[ /etc/cont-init.d/50-configure_pulseaudio.sh: executing... ]
**** Configure pulseaudio ****
  - Enable pulseaudio service.
  - Configure pulseaudio to pipe audio to a socket
DONE

[ /etc/cont-init.d/60-configure_gpu_driver.sh: executing... ]
**** Found Intel device '12th Gen Intel(R) Core(TM) i5-12500T 12th Gen Intel(R) Core(TM) i5-12500T To Be Filled By O.E.M. CPU @ 1.9GHz' ****
  - Mesa has already been installed into this container
**** No AMD device found ****
**** Found NVIDIA device 'NVIDIA GeForce RTX 4070' ****
  - Downloading driver v545.23.08`

[alansari]

v545.23.08 is not hosted on https://download.nvidia.com/XFree86/Linux-x86_64 though thats what is offered on truenas. You wil need to get the run file and rename it to NVIDIA_545.23.08.run and place it in your dataset you mounted for /home/default/Downloads/. Or you could downgrade truenas to a previous snapshot/version from the grub boot manager. Or jump the gun to electric eel and migrate to useing native docker compose to spin up the container.
edit: fingers crossed today's schedualed maitenence release includes a diff nvidia driver release

[Cribz679]

Amazing. Thank you. That worked. :)

[alansari]

Just an FYI, i bit the bullet and updated my truenas installation to latest.
In order to get the container up and running i used these steps:

• wget https://developer.download.nvidia.com/compute/cuda/12.3.2/local_installers/cuda_12.3.2_545.23.08_linux.run
• extract the aformentioned run file to get the driver installer's .run with --extract flag
• Copied and renamed the .run from whereever you extracted in the last step to the /home/default/NVIDIA_545.23.08.run
• Profit!

edit: I would highly recomend moving away from the helm/truecharts version to jailmaker/nspawn implementation. Detailed walkthrough can be found in a thread on the steam-headless discord server.

[seisdr]

Just an FYI, i bit the bullet and updated my truenas installation to latest. In order to get the container up and running i used these steps:

• wget https://developer.download.nvidia.com/compute/cuda/12.3.2/local_installers/cuda_12.3.2_545.23.08_linux.run
• extract the aformentioned run file to get the driver installer's .run with --extract flag
• Copied and renamed the .run from whereever you extracted in the last step to the /home/default/NVIDIA_545.23.08.run
• Profit!

edit: I would highly recomend moving away from the helm/truecharts version to jailmaker/nspawn implementation. Detailed walkthrough can be found in a thread on the steam-headless discord server.

home/default or home/defaults/Downloads?
Regardless it doesn't work!

[WillKirkmanM]

This has been fixed due to the merged pull request #153. Can Close.