Ubuntu16.04下安裝nvidia-docker2

若docker-ce、nvidia、CUDA等都安裝完成之后，開啟docker服務時，能夠正常運行，並有預測結果，那表示服務開啟沒問題；若都安裝成功之后，用docker命令開啟服務時，一直報錯，可能表示你沒有安裝nvidia-docker2：
報錯信息：

tf-serving@tfserving-KVM:~/model/yolo$ docker: Error response from daemon: OCI runtime create failed: unable to retrieve OCI runtime error (open /run/containerd/io.containerd.runtime.v1.linux/moby/f40efd9bd62fc00e53e1d48ea0cbbf8e2c76efdac28238239c6a0c49f52aaebc/log.json: no such file or directory): fork/exec /usr/bin/nvidia-container-runtime: no such file or directory: : unknown.

$ systemctl status docker.service

結果：
docker.service - Docker Application Container Engine
   Loaded: loaded (/lib/systemd/system/docker.service; enabled; vendor preset: enabled)
  Drop-In: /etc/systemd/system/docker.service.d
           └─override.conf
   Active: failed (Result: start-limit-hit) since 四 2019-06-20 11:43:26 CST; 8s ago
     Docs: https://docs.docker.com
  Process: 8024 ExecStart=/usr/bin/dockerd --host=fd:// --add-runtime=nvidia=/usr/bin/nvidia-container-runtime (code=exited, status=1/FAILURE)
 Main PID: 8024 (code=exited, status=1/FAILURE)

6月 20 11:43:26 tfserving-KVM systemd[1]: Failed to start Docker Application Container Engine.
6月 20 11:43:26 tfserving-KVM systemd[1]: docker.service: Unit entered failed state.
6月 20 11:43:26 tfserving-KVM systemd[1]: docker.service: Failed with result 'exit-code'.
6月 20 11:43:26 tfserving-KVM systemd[1]: docker.service: Service hold-off time over, scheduling restart.
6月 20 11:43:26 tfserving-KVM systemd[1]: Stopped Docker Application Container Engine.
6月 20 11:43:26 tfserving-KVM systemd[1]: docker.service: Start request repeated too quickly.
6月 20 11:43:26 tfserving-KVM systemd[1]: Failed to start Docker Application Container Engine.
6月 20 11:43:26 tfserving-KVM systemd[1]: docker.service: Unit entered failed state.
6月 20 11:43:26 tfserving-KVM systemd[1]: docker.service: Failed with result 'start-limit-hit'.

出現類似這樣的信息，錯誤，解決辦法，可嘗試：

（1）重新安裝docker-ce及tensorflow_model_server;

（2）安裝nvidia-docker2，步驟如下：

0 If you have nvidia-docker 1.0 installed: we need to remove it and all existing GPU containers

           docker volume ls -q -f driver=nvidia-docker | xargs -r -I{} -n1 docker ps -q -a -f volume={} | xargs -r docker rm - 
          f
sudo apt-get purge -y nvidia-docker 
         

 

           1. 創建用戶組 
          
sudo groupadd docker 
         

 

           2.添加用戶進入docker組，以便該用戶可以使用docker 
          
sudo gpasswd -a ${USER} docker 
         

 

           3.重啟服務 
          
sudo service docker restart 
         

 

           4.修改docker倉庫和存放目錄 
          
sudo vim /etc/docker/daemon.json 
         

 

5.修改如下，其中https://cwoel6s9.mirror.aliyuncs.com 為從阿里服務器申請的鏡像地址，/data/docker     為存放目錄

           { 
          
    "registry-mirrors": [ 
          
        "https://cwoel6s9.mirror.aliyuncs.com" 
          
    ], 
          
    "graph":"/data/docker", 
          
    "storage-driver": "overlay", 
          
    "runtimes": { 
          
        "nvidia": { 
          
            "path": "nvidia-container-runtime", 
          
            "runtimeArgs": [] 
          
        } 
          
    } 
          
}  
         

 

           6.重啟docker，並使用docker info查看修改信息 
          
systemctl daemon-reload 
          
systemctl restart docker 
          
docker info 
         

 

           7.nvidia-docker 安裝,添加源 
          
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | 
           \
  sudo apt-key add - 
         

 

 
          curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | \
  sudo tee /etc/apt/sources.list.d/nvidia-docker.list
 
         

 

sudo apt-get update

 

           8.安裝nvidia-docker2軟件包並重新加載Docker守護程序配置 
          
sudo apt-get install nvidia-docker2 
          
sudo pkill -SIGHUP dockerd 
         

 

安裝完成后，當需要使用gpu時，使用nvidia-docker代替docker！

# 注意：執行這一步，如果沒有安裝nvidia驅動，會報錯：
# docker：ERROR response from Daemon....
# 安裝nvida驅動就可以了