ansible-role-nvidia

The role deploys the NVIDIA driver in GPU machines, these can be baremetal, VMs or Dockers.

Supported Operating systems are Centos7 and Ubuntu 16.04.

Requirements

The machine, VM or docker should have a NVIDIA GPU, can be checked with:

lspci |grep NVIDIA

Role Variables

These variables should be changed depending on your type of machine. Specifically the driver version may need to be set if you are using docker containers where the version on the docker image has to match the version on the host machine:

• nvidia_driver_version: NVIDIA driver version, check the default value in defaults/main.yml as well as a list of valid versions
• kmod_install: true|false

The variable kmod_install enables the installation (or not) of the nvidia kernel module. In the case of docker images the value should be false, while in the case of baremetal or VMs it should be true.

These variables have pre-configured values that should not be needed to change:

• base_url: Base URL of NVIDIA download
• nvidia_driver: Filename of the NVIDIA driver, depends on the variable nvidia_driver_version
• nvidia_url: Full URL of the NVIDIA driver, depends on the variables base_url and nvidia_driver

Example Playbook

Playbook example: deployment of driver version 352.93 on a VM

- hosts: servers
  roles:
    - { role: LIP-Computing.ansible-role-nvidia, nvidia_driver_version: 352.93, kmod_install: true }

An example of deployment of driver 352.99 on a docker container:

- hosts: servers
  roles:
    - { role: LIP-Computing.ansible-role-nvidia, nvidia_driver_version: 352.99, kmod_install: false }

License

Apache v2

Author Information

Mario David: [email protected]

LIP Lisbon: http://www.lip.pt

Indigo DataCloud: https://www.indigo-datacloud.eu/