centos7 安装 NVIDIA Docker

安装环境：

1、centos7.3

2、NVIDIA Corporation GP106 [GeForce GTX 1060 6GB]

安装nvidia-docker 

a、安装docker 可参考centos7 安装docker

b：

# Install nvidia-docker and nvidia-docker-plugin
wget -P /tmp https://github.com/NVIDIA/nvidia-docker/releases/download/v1.0.1/nvidia-docker-1.0.1-1.x86_64.rpm
sudo rpm -i /tmp/nvidia-docker*.rpm && rm /tmp/nvidia-docker*.rpm
sudo systemctl start nvidia-docker

# Test nvidia-smi
nvidia-docker run --rm nvidia/cuda nvidia-smi

如果出现异常：

[root@localhost ~]# nvidia-docker run --rm nvidia/cuda nvidia-smi
/usr/bin/docker-current: Error response from daemon: create nvidia_driver_384.69: create nvidia_driver_384.69: Error looking up volume plugin nvidia-docker: plugin not found.
See '/usr/bin/docker-current run --help'.

则检查nvidia-docker是否启动：

[root@localhost ~]# systemctl status nvidia-docker
● nvidia-docker.service - NVIDIA Docker plugin
   Loaded: loaded (/usr/lib/systemd/system/nvidia-docker.service; disabled; vendor preset: disabled)
   Active: inactive (dead)
     Docs: https://github.com/NVIDIA/nvidia-docker/wiki
[root@localhost ~]# systemctl start nvidia-docker 
[root@localhost ~]# systemctl status nvidia-docker
● nvidia-docker.service - NVIDIA Docker plugin
   Loaded: loaded (/usr/lib/systemd/system/nvidia-docker.service; disabled; vendor preset: disabled)
   Active: active (running) since Wed 2017-08-30 03:18:32 CST; 5s ago
     Docs: https://github.com/NVIDIA/nvidia-docker/wiki
  Process: 11135 ExecStartPost=/bin/sh -c /bin/echo unix://$SOCK_DIR/nvidia-docker.sock > $SPEC_FILE (code=exited, status=0/SUCCESS)
  Process: 11131 ExecStartPost=/bin/sh -c /bin/mkdir -p $( dirname $SPEC_FILE ) (code=exited, status=0/SUCCESS)
 Main PID: 11130 (nvidia-docker-p)
   CGroup: /system.slice/nvidia-docker.service
           └─11130 /usr/bin/nvidia-docker-plugin -s /var/lib/nvidia-docker

Aug 30 03:18:32 localhost.localdomain systemd[1]: Starting NVIDIA Docker plugin...
Aug 30 03:18:32 localhost.localdomain systemd[1]: Started NVIDIA Docker plugin.
Aug 30 03:18:32 localhost.localdomain nvidia-docker-plugin[11130]: /usr/bin/nvidia-docker-plugin | 2017/08/30 03:18:32 Loading NV...mory
Aug 30 03:18:32 localhost.localdomain nvidia-docker-plugin[11130]: /usr/bin/nvidia-docker-plugin | 2017/08/30 03:18:32 Loading NV...rary
Aug 30 03:18:33 localhost.localdomain nvidia-docker-plugin[11130]: /usr/bin/nvidia-docker-plugin | 2017/08/30 03:18:33 Discoverin...ices
Aug 30 03:18:33 localhost.localdomain nvidia-docker-plugin[11130]: /usr/bin/nvidia-docker-plugin | 2017/08/30 03:18:33 Provisioni...umes
Aug 30 03:18:33 localhost.localdomain nvidia-docker-plugin[11130]: /usr/bin/nvidia-docker-plugin | 2017/08/30 03:18:33 Serving pl...cker
Aug 30 03:18:33 localhost.localdomain nvidia-docker-plugin[11130]: /usr/bin/nvidia-docker-plugin | 2017/08/30 03:18:33 Serving re...3476
Hint: Some lines were ellipsized, use -l to show in full.
[root@localhost ~]# nvidia-docker run --rm nvidia/cuda nvidia-smi
Tue Aug 29 19:18:46 2017       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.69                 Driver Version: 384.69                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 106...  Off  | 00000000:01:00.0 Off |                  N/A |
| 43%   39C    P0    22W / 120W |     10MiB /  6072MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

如果出现以下情况：

[root@localhost ~]# nvidia-docker run --rm nvidia/cuda nvidia-smi
NVIDIA-SMI couldn't find libnvidia-ml.so library in your system. Please make sure that the NVIDIA Display Driver is properly installed and present in your system.
Please also try adding directory that contains libnvidia-ml.so to your system PATH.key

可以尝试一下加参数--privileged=true ，如果没问题，则需要修改selinux配置,编辑/etc/selinux/config：

SELINUX=disabled 
SELINUXTYPE=targeted

或者尝试selinux模式为permissive模式 setenforce 0

重启机器，重启服务器即可。

可参考：

https://github.com/NVIDIA/nvidia-docker/issues/407

https://github.com/NVIDIA/nvidia-docker