Ubuntu离线安装nvidia-docker完整过程（最简单的解决方法解决nvidia-docker: command not found/Unknown runtime specified）

问题说明

安装完docker、NVIDIA驱动后，执行指令：

nvidia-docker

报错如下：

nvidia-docker: command not found

第二种错误：

Error response from daemon: Unknown runtime specified nvidia.
See 'docker run --help'.

【注】第二种错误的解决方法直接看 【3.3修改配置文件 daemon.json】 再按照4,5步骤依次进行

运行环境

• Ubuntu 18.04
• Docker version 18.09.6
• NVIDIA-SMI 460.32.03 Driver Version: 460.32.03 CUDA Version: 11.2

解决方法

• 1、下载5个deb文件
  libnvidia-container1
  libnvidia-container-tools
  nvidia-container-toolkit
  nvidia-container-runtime
  nvidia-docker2

  docker官网下载链接：
  http://mirror.cs.uchicago.edu/nvidia-docker/libnvidia-container/stable/ubuntu16.04/amd64/

  我下载的文件：

  libnvidia-container-tools_1.7.0-1_amd64.deb  
  nvidia-container-toolkit_1.7.0-1_amd64.deb
  libnvidia-container1_1.7.0-1_amd64.deb  
  nvidia-container-runtime_3.7.0-1_all.deb     
  nvidia-docker2_2.8.0-1_all.deb
  

• 2、安装

  执行命令：

  sudo dpkg -i ./lib*  ./nvidia*
  

• 3、如果第二步安装过程提示下面的错误：

  dpkg: dependency problems prevent configuration of nvidia-docker2:
   nvidia-docker2 depends on docker-ce (>= 18.06.0~ce~3-0~ubuntu) | docker-ee (>= 18.06.0~ce~3-0~ubuntu) | docker.io (>= 18.06.0); however:
    Package docker-ce is not installed.
    Package docker-ee is not installed.
    Package docker.io is not installed.
  
  

  解决方法：安装 docker-ce，docker-ee， docker.io

• 3.1下载下面文件
  【说明】我安装的 docker 版本是 18.09.6。（docker 20.10.2的相关文件见后面）

  	containerd.io_1.2.6-3_amd64.deb  
  	docker-ce_18.09.6~3-0~ubuntu-bionic_amd64.deb  
  	docker-ce-cli_18.09.6~3-0~ubuntu-bionic_amd64.deb
  

  官网下载地址：https://download.docker.com/linux/ubuntu/dists/bionic/pool/stable/amd64/

  我的下载地址：

  https://download.docker.com/linux/ubuntu/dists/bionic/pool/stable/amd64/docker-ce_18.09.6~3-0~ubuntu-bionic_amd64.deb
  https://download.docker.com/linux/ubuntu/dists/bionic/pool/stable/amd64/docker-ce-cli_18.09.6~3-0~ubuntu-bionic_amd64.deb
  https://download.docker.com/linux/ubuntu/dists/bionic/pool/stable/amd64/containerd.io_1.2.6-3_amd64.deb
  

  下载文件保存到文件夹：docker_deb/

• 3.2安装

  	cd docker_deb/
  	sudo dpkg -i ./*deb
  

    安装末尾提示：
  

  	Configuration file '/etc/docker/daemon.json'
  	 ==> File on system created by you or by a script.
  	 ==> File also in package provided by package maintainer.
  	   What would you like to do about it ?  Your options are:
  	    Y or I  : install the package maintainer's version
  	    N or O  : keep your currently-installed version
  	      D     : show the differences between the versions
  	      Z     : start a shell to examine the situation
  	 The default action is to keep your current version.
  	*** daemon.json (Y/I/N/O/D/Z) [default=N] ? 
  	
  

  选择：N

• 3.3修改配置文件 daemon.json
  首先查看 daemon.json：

  {
    "data-root": "/var/lib/docker",
    "exec-opts": ["native.cgroupdriver=systemd"],
    "insecure-registries": ["xxx"],
    "max-concurrent-downloads": 10,
    "live-restore": true,
    "log-driver": "json-file",
    "log-level": "warn",
    "log-opts": {
      "max-size": "50m",
      "max-file": "1"
      },
    "storage-driver": "overlay2"
  }
  

  问题：缺少 runtimes、default-runtime，新增如下内容：

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

• 4、重启docker服务
  先停止所有运行的容器：

  docker stop $(docker ps -a -q)
  

  再重启 docker 服务：

  sudo systemctl daemon-reload
  sudo systemctl restart docker
  

• 5、验证nvidia-docker

  执行指令：

  nvidia-docker -v
  

  返回结果：

  Docker version 18.09.6, build 481bc77
  

  说明 nvidia-docker 安装成功

测试nvidia-docker

nvidia-docker  run -it -d \
    --name your_name \
    -e TZ='Asia/Shanghai' \
    -d your_ai_image:latest

不报错则说明 nvidia-docker 正常可用。

docker 20.10.2的安装

下载文件：

containerd.io_1.4.6-1_amd64.deb 
docker-ce_20.10.2_3-0_ubuntu-bionic_amd64.deb  
docker-ce-cli_20.10.2_3-0_ubuntu-bionic_amd64.deb

对应的下载链接：

https://download.docker.com/linux/ubuntu/dists/bionic/pool/stable/amd64/containerd.io_1.4.6-1_amd64.deb
https://download.docker.com/linux/ubuntu/dists/bionic/pool/stable/amd64/docker-ce-cli_20.10.2~3-0~ubuntu-bionic_amd64.deb
https://download.docker.com/linux/ubuntu/dists/bionic/pool/stable/amd64/docker-ce_20.10.2~3-0~ubuntu-bionic_amd64.deb

其他过程同 docker 18.09.6 一致。

参考

【1】https://blog.csdn.net/zengNLP/article/details/126732645?spm=1001.2014.3001.5502
【2】http://mirror.cs.uchicago.edu/nvidia-docker/libnvidia-container/stable/ubuntu16.04/amd64/
【3】https://download.docker.com/linux/ubuntu/dists/bionic/pool/stable/amd64/
【4】https://www.codenong.com/cs109532661/