[Docker] nvidia container runtime 설정

nvidia container runtime 설정

nvidia container runtime 을 설정하기 위해서는 꼭 docker-ce 로 docker 가 설치가 되어 있어야 됩니다.
(일반 RHEL/CentOS에서 제공되는 docker package로는 설치 불가)

docker-ce 설치

• yum-utils 설치

  $ yum -y install yum-utils
  

• docker-ce Repository 연결

  $ yum-config-manager \ 
  >     --add-repo \
  >     https://download.docker.com/linux/centos/docker-ce.repo
  Loaded plugins: fastestmirror 
  adding repo from: https://download.docker.com/linux/centos/docker-ce.repo 
  grabbing file https://download.docker.com/linux/centos/docker-ce.repo to /etc/yum.repos.d/docker-ce.repo 
  repo saved to /etc/yum.repos.d/docker-ce.repo 
  $ 
  

• docker-ce 설치

  $ yum install docker-ce
  

• docker 서비스 시작

  $ systemctl enable docker
  Created symlink from /etc/systemd/system/multi-user.target.wants/docker.service to /usr/lib/systemd/system/docker.service. 
  $ systemctl start docker
  

• docker 기본 runtime 확인

  # docker info 
  ... 중략
   Runtimes: runc
   Default Runtime: runc
   Init Binary: docker-init
   containerd version: 894b81a4b802e4eb2a91d1ce216b8817763c29fb
   runc version: 425e105d5a03fabd737a126ad93d62a9eeede87f
   init version: fec3683
  ... 중략
  

  위와 같이 Default 로 지정된 runtime 은 runc 입니다.

nvidia container runtime 설치

• nvidia container runtime repository 연결

  $ distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
  $ curl -s -L https://nvidia.github.io/nvidia-container-runtime/$distribution/nvidia-container-runtime.repo | \
  sudo tee /etc/yum.repos.d/nvidia-container-runtime.repo
  

  참고문서 : https://nvidia.github.io/nvidia-container-runtime/

• nvidia container runtime 설치

  $ yum install nvidia-container-runtime
  

• Daemon configuration file 수정 및 systemd 수정
  • Daemon configuration file 수정

    $ vi /etc/docker/daemon.json
    {
      "default-runtime": "nvidia",
      "runtimes": {
          "nvidia": {
              "path": "/usr/bin/nvidia-container-runtime",
              "runtimeArgs": []
       }
      }
    }
    

  • systemd 수정

    $ mkdir -p /etc/systemd/system/docker.service.d
    $ tee /etc/systemd/system/docker.service.d/override.conf <<EOF
    [Service]
    ExecStart=
    ExecStart=/usr/bin/dockerd --host=fd:// --add-runtime=nvidia=/usr/bin/nvidia-container-runtime
    EOF
    $ systemctl daemon-reload
    $ systemctl restart docker
    

• nvidia container runtime 설치 확인

  $ docker info 
  ... 중략
   Runtimes: nvidia runc
   Default Runtime: nvidia
   Init Binary: docker-init
   containerd version: 894b81a4b802e4eb2a91d1ce216b8817763c29fb
   runc version: 425e105d5a03fabd737a126ad93d62a9eeede87f
   init version: fec3683
  ... 중략
  

  위와 같이 사용가능한 docker runtime 은 nvidia, runc 이고 Default runtime 은 nvidia 입니다.
  앞으로 docker run 명령을 통해 생성되는 container 는 nvidia-container-runtime 을 이용하여 생성될 것입니다.

마치며

상기 내용은 NVIDIA DGX Station 시스템에 Red Hat 운영체제 설치하고 CUDA 설정을 하면서 경험한 내용입니다.

[root@localhost sosreport-test-2019-xx-xx-xxxxxx]# cat proc/driver/nvidia/gpus/*/information
Model:       Tesla V100-DGXS-16GB
IRQ:         144
GPU UUID:    GPU-11111111-1111-1111-1111-111111111111
Video BIOS:      88.00.24.00.01
Bus Type:    PCIe
DMA Size:    47 bits
DMA Mask:    0x7fffffffffff
Bus Location:    0000:07:00.0
Device Minor:    0
Blacklisted:     No
Model:       Tesla V100-DGXS-16GB
IRQ:         145
GPU UUID:    GPU-22222222-2222-2222-2222-222222222222
Video BIOS:      88.00.24.00.01
Bus Type:    PCIe
DMA Size:    47 bits
DMA Mask:    0x7fffffffffff
Bus Location:    0000:08:00.0
Device Minor:    1
Blacklisted:     No
Model:       Tesla V100-DGXS-16GB
IRQ:         146
GPU UUID:    GPU-33333333-3333-3333-3333-333333333333
Video BIOS:      88.00.24.00.01
Bus Type:    PCIe
DMA Size:    47 bits
DMA Mask:    0x7fffffffffff
Bus Location:    0000:0e:00.0
Device Minor:    2
Blacklisted:     No
Model:       Tesla V100-DGXS-16GB
IRQ:         147
GPU UUID:    GPU-44444444-4444-4444-4444-444444444444
Video BIOS:      88.00.24.00.01
Bus Type:    PCIe
DMA Size:    47 bits
DMA Mask:    0x7fffffffffff
Bus Location:    0000:0f:00.0
Device Minor:    3
Blacklisted:     No

Tesla V100 이 무려 4장이나 설치된 어마어마한 장비였습니다……. :(

참고 문서

• NVIDIA DGX Station : https://www.nvidia.com/ko-kr/data-center/dgx-station/
• NVIDIA Repository : https://nvidia.github.io/nvidia-container-runtime/
• Github nvidia-container-runtime : https://github.com/NVIDIA/nvidia-container-runtime
• Install docker-ce : https://docs.docker.com/install/linux/docker-ce/centos/