NVIDIA Container Toolkit - installation Error on OMV6

Hello,

i tried to install the Toolkit or Drivers for the Docker from Nvidia, i found this Guide here

Post

RE: [HowTo] Nvidia hardware transcoding on OMV 5 in a Plex docker container

******************************* OMV6 **************************************

For OMV 7 proceed to OMV 7

If you are having issues with the new docker-compose-Plugin, read the comments regarding that issue.

Tested and running on:
- Linux Debian 11, Kernel 5.18.0-0.bpo.4-amd64, 5.18.0-0.deb11.4-amd64, 5.19.0-0.deb11.2-amd64, 5.19.17-1-pve, Linux 6.1.15-1-pve, Linux 6.2.16-11-bpo11-pve, 6.2.16-20-pve, 6.5.13-1-pve, 6.1.15-1-pve, 6.8.12-6-pve, 6.11.0-2-pve, 6.11.11-1-pve
- 6.9.1-1 (Shaitan), 7.0.1-1…

chris_kmn

May 21st 2022

But after Step3 i have the problem that i get no graphics card displayed with "nvidia-smi" command, i only get the error

"NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running."

i dont know what to do? has please someone a hint?

mY sYTEM

Version
6.0.5-3 (Shaitan)

Prozessor
AMD Ryzen 7 3700X 8-Core Processor

Kernel
Linux 5.14.0-0.bpo.2-amd64

Graphics Card

Nvidia GTX 1650 OC 4GB

Quote from BernH

I just went through this with my buddy. It took 4 attempts to get it all in a working. Each time I followed the same steps but it just seems like its a bit hit and miss, so I's suggest you purge all nvidia stuff and try agin until you get a working setup.

Here is a list of my process (written from memory so please forgive any errors)

Code

apt-get purge *nvidia*
apt-get update
apt-get upgrade
apt-get install linux-headers-amd64
apt-get install nvidia-detect
nvidia-detect (to verify if card is detected correctly)
apt-get install nvidia-driver linux-image-amd64
reboot
apt-get install nvidia-smi
nvidia-smi (to verify if nvidi-smi is able to access the card)
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list && sudo apt-get update
apt-get install -y nvidia-container-toolkit
nvidia-ctk runtime configure --runtime=docker
nvidia-ctk runtime configure --runtime=containerd

restart docker from GUI

Display More

hi

thx for your answer i give it a try, you change nothing on the two files?

/etc/docker/daemon.json

and

/etc/nvidia-container-runtime/config.toml

?

EDIT

nvidia-detect

gives me this, he found the card

nvidia-smi

gives me this, i think its okay?

but there stands : CUDA Version: N/A

i think thats not right

and unter portainer, i dont have the nvida runtime

Quote from BernH

There is an nvidia cuda driver that needs to be installed after the rest is running. Sorry i forgot that one. I can't remember the name of it off the top of my head. I just did and apt search nvidia and looked for it. Then an apt-get install <name of cuda driver package>

I didn't change the toml file of the docker json manually. the nvidia-ctk command does that for the docker.json.

the nvidia runtime is injected by that command. you will also then need the nvidia docker toolkit in order for docker to use the card.

This was my first time doing an nvidia card. I personally have an intel arc 380. It is much easier to set up and gives better quality than nvidia.

Display More

okay i think it works now? under Enviroment now i have th eoption "NVIDIA"

Quote from BernH

yep, that looks like it should be functioning

yeah thats works well, i tested it.

wow thank you