Docker Error - Unknown or Invalid Runtime Name: Nvidia

[YuminosukeSato]

1. Issue or feature description

I'm experiencing a runtime error with Docker when using the Nvidia runtime. Despite having a successful initial Docker command and editing the Docker configuration, I encounter an error when executing a script.

2. Steps to reproduce the issue

1. Initial Docker Command:

   sudo docker run --rm --runtime=nvidia --gpus all ubuntu nvidia-smi

   This command executes successfully.

2. Edit Docker Configuration:
   Edited /etc/docker/daemon.json to set Nvidia as the default runtime:

   {
       "runtimes": {
           "nvidia": {
               "path": "nvidia-container-runtime",
               "runtimeArgs": []
           }
       },
       "default-runtime": "nvidia"
   }

3. Execute Script with Modified Settings:
   Running the script scripts/run_dev.sh ~/workspaces/isaac_ros-dev/ results in an error:

   docker: Error response from daemon: unknown or invalid runtime name: nvidia.
   

3. Information to attach (optional if deemed irrelevant)

• Output of nvidia-container-cli -k -d /dev/tty info
• Kernel version from uname -a
• Relevant kernel output lines from dmesg
• Driver information from nvidia-smi -a
• Docker version from docker version 24.0.7, build afdd53b
• NVIDIA packages version from dpkg -l '*nvidia*' or rpm -qa '*nvidia*'
• NVIDIA container library version from nvidia-container-cli -V
• NVIDIA container library logs (see troubleshooting)
• Docker command, image and tag used
• Additional information:
  • Operating System: Ubuntu 22.04
  • Output of sudo docker info | grep -i runtime shows Runtimes: runc io.containerd.runc.v2 nvidia and Default Runtime: nvidia
  • Confirmation of successful installation of the NVIDIA Container Toolkit

[elezar]

Just to make sure that this has been done, have you restarted the docker daemon after making the configuration changes:

sudo systemctl restart docker

(I would expect so since the runtime shows up in the docker info output).

[EmmanuelMr18]

Are you using Ubuntu with WSL?
If so, I found that you need to specify the runtimes in the configuration file of your docker desktop on windows.
This looks that fixed my problem, now running my image with docker is working.

https://stackoverflow.com/questions/77323535/add-nvidia-runtime-to-docker-runtimes-on-windows-wsl

[wirytiox]

for anyone who wanted to copy paste it:

{
  "builder": {
    "gc": {
      "defaultKeepStorage": "20GB",
      "enabled": true
    }
  },
  "experimental": false,
  "runtimes": {
    "nvidia": {
      "path": "/usr/bin/nvidia-container-runtime",
      "runtimeArgs": []
    }
  }
}