/Docker-Nvidia-Container-Toolkit

Docker Nvidia Container Toolkit

MIT LicenseMIT

Docker with the Nvidia Container Toolkit

Table of contents

  1. Notice
  2. Summarized environments about the Docker with the Nvidia Container Toolkit
  3. How to install the docker
  4. How to install the Nvidia Container Toolkit
  5. Docker image
  6. Docker container
  7. Reference

1. Notice

  • A guide for Docker with the Nvidia Container Toolkit (i.e. Nvidia-Docker 2.0)
  • I recommend that you should ignore the commented instructions with an octothorpe, #.
  • Modified date: Dec. 22, 2020.

2. Summarized environments about the Docker with the Nvidia Container Toolkit

  • Operating System (OS): Ubuntu MATE 18.04.3 LTS (Bionic)
  • Graphics Processing Unit (GPU): NVIDIA TITAN Xp, 1ea
  • GPU driver: Nvidia-440.100
  • CUDA toolkit: CUDA 10.2
  • cuDNN: cuDNN v7.6.5
  • Docker: Docker version 20.10.1, build 831ebea
  • Docker image: nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04

3. How to install the docker

  1. Docker installation.

    usrname@hostname:~/curr_path$ curl https://get.docker.com | sh
usrname@hostname:~/curr_path$ sudo systemctl start docker && sudo systemctl enable docker

  2. Check the docker version.

    usrname@hostname:~/curr_path$ sudo docker --version
    Docker version 20.10.1, build 831ebea

4. How to install the Nvidia Container Toolkit

  1. Setup the repository and the Gnu Privacy Guard (GPU) key.

    usrname@hostname:~/curr_path$ distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
usrname@hostname:~/curr_path$ curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
usrname@hostname:~/curr_path$ curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
usrname@hostname:~/curr_path$ sudo apt-get update

  2. Install the Nvidia-Docker 2.0 for the Nvidia Container Toolkit.

    usrname@hostname:~/curr_path$ sudo apt-get install -y nvidia-docker2
usrname@hostname:~/curr_path$ sudo systemctl restart docker

  3. Test the Nviida-Docker 2.0.

    usrname@hostname:~/curr_path$ sudo docker run --rm --gpus all nvidia/cuda:10.2-base nvidia-smi
    Tue Dec 22 08:40:19 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.100      Driver Version: 440.100      CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  TITAN Xp            Off  | 00000000:01:00.0  On |                  N/A |
| 23%   30C    P8    10W / 250W |    518MiB / 12192MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                            
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

5. Docker image

  1. List docker images.

    usrname@hostname:~/curr_path$ sudo docker images
    REPOSITORY    TAG                             IMAGE ID       CREATED        SIZE
nvidia/cuda   10.2-base                       fbd4a873bb90   7 days ago     107MB

  2. Pull a docker image.

    • Target image: nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04 
      usrname@hostname:~/curr_path$ sudo docker pull nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04

  3. Remove a docker image.

    • Target image tag: nvidia/cuda:10.2-base
    • Target image ID: fbd4a873bb90 
      usrname@hostname:~/curr_path$ sudo docker rmi fbd4a873bb90

  4. Remove a docker image with the corresponding docker container.

    • Target image tag: nvidia/cuda:10.2-base
    • Target image ID: fbd4a873bb90 
      usrname@hostname:~/curr_path$ sudo docker rmi -f fbd4a873bb90

6. Docker container

  1. Run a docker container.

    • There are two options for command: i) image tag, ii) image ID.
    • Target image tag: nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04
    • Target image ID: a4bdb1190443 
      # i) image tag
usrname@hostname:~/curr_path$ sudo docker run -it nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04 /bin/bash
      # ii) iamge ID
usrname@hostname:~/curr_path$ sudo docker run -it fbd4a873bb90 /bin/bash

  2. List docker containers.

    • List working docker containers
      • There are two options for result: i) image tag, ii) image ID. 
        usrname@hostname:~/curr_path$ sudo docker ps
        # i) image tag
CONTAINER ID   IMAGE                                       COMMAND       CREATED         STATUS         PORTS     NAMES
2170c31af220   nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04   "/bin/bash"   3 minutes ago   Up 3 minutes             vigilant_feynman
        # ii) image ID
CONTAINER ID   IMAGE          COMMAND       CREATED         STATUS         PORTS     NAMES
f7912486a8dc   fbd4a873bb90   "/bin/bash"   5 minutes ago   Up 5 minutes             distracted_buck
    • List all docker containers
      • Please note that the results are not been shown.
      • There are two options for result: i) image tag, ii) image ID. 
        usrname@hostname:~/curr_path$ sudo docker ps --all

  3. Pause and unpause a docker container.

    • Target container ID: f7912486a8dc
    • Pause 
      usrname@hostname:~/curr_path$ sudo docker pause f7912486a8dc
    • Unpause 
      usrname@hostname:~/curr_path$ sudo docker unpause f7912486a8dc

  4. Stop a docker container.

    • Target container ID: f7912486a8dc 
      usrname@hostname:~/curr_path$ sudo docker stop f7912486a8dc

  5. Start the stopped docker container.

    • Get the stopped docker container ID. 
      usrname@hostname:~/curr_path$ sudo docker ps -all
      CONTAINER ID   IMAGE          COMMAND       CREATED          STATUS                       PORTS     NAMES
f7912486a8dc   fbd4a873bb90   "/bin/bash"   23 minutes ago   Exited (127) 2 minutes ago             distracted_buck
    • Target container ID: f7912486a8dc 
      usrname@hostname:~/curr_path$ sudo docker start f7912486a8dc
    • Check whether the docker container is started. 
      usrname@hostname:~/curr_path$ sudo docker ps
      CONTAINER ID   IMAGE          COMMAND       CREATED          STATUS          PORTS     NAMES
f7912486a8dc   fbd4a873bb90   "/bin/bash"   27 minutes ago   Up 2 seconds              distracted_buck

  6. Restart the docker container.

    • Please note that the results are not been shown.
    • Target container ID: f7912486a8dc 
      usrname@hostname:~/curr_path$ sudo docker restart f7912486a8dc

  7. Attach a working docker container.

    • Target container ID: f7912486a8dc 
      usrname@hostname:~/curr_path$ sudo docker attach f7912486a8dc
      root@f7912486a8dc:/#

  8. Remove a docker container.

    • Please note that the docker container should be stopped before removing it.
    • Target container ID: f7912486a8dc 
      usrname@hostname:~/curr_path$ sudo docker rm fbd4a873bb90
    • Check whether the docker container is removed. 
      usrname@hostname:~/curr_path$ sudo docker start
      Error response from daemon: No such container: f7912486a8dc
Error: failed to start containers: f7912486a8dc

  9. Remove all docker containers.

    • Please note that the docker containers should be stopped before removing them. 
      sudo docker rm `sudo docker ps -a -q`
    • Check whether the docker containers are removed. 
      usrname@hostname:~/curr_path$ sudo docker ps --all
      CONTAINER ID   IMAGE     COMMAND   CREATED   STATUS    PORTS     NAMES

7. Reference

  1. Docker.
  2. Github for the Nvidia-Docker.
  3. Document for the Nvidia-Docker.