main.yml

---
- name: "Check GPU card info with lspci"
  raw: "lspci | grep -E '[0-9a-fA-F][0-9a-fA-F]:[0-9a-fA-F][0-9a-fA-F].[0-9] (3D|VGA compatible) controller: NVIDIA Corporation.*'"
  register: lspci_grep
  failed_when: false
  changed_when: false
  check_mode: false
  environment: {}

- name: "Check NVIDIA GPU exits or not"
  set_fact:
    unmet_requirements: "{{ unmet_requirements + ['NVIDIA GPU card is not detected'] }}"
  changed_when: false
  check_mode: false
  environment: {}
  when:
    - lspci_grep.rc != 0

- name: "Check NVIDIA GPU drivers with nvidia-smi"
  raw: "nvidia-smi"
  register: nvidia_smi
  failed_when: false
  changed_when: false
  check_mode: false
  environment: {}

- name: "Check NVIDIA GPU drivers is installed or not "
  set_fact:
    unmet_requirements: "{{ unmet_requirements + ['NVIDIA GPU drivers is not detected'] }}"
  changed_when: false
  check_mode: false
  environment: {}
  when:
    - nvidia_smi.rc != 0

- name: "Check NVIDIA GPU card number with nvidia-smi"
  raw: "nvidia-smi --query-gpu=gpu_name --format=csv | tail --lines=+2 | wc -l"
  register: nvidia_gpu_count
  failed_when: false
  changed_when: false
  check_mode: false
  environment: {}
  when:
    # if nvidia-smi doesn't work, we can skip this step
    - nvidia_smi.rc == 0 

- name: "Check NVIDIA GPU card number is matched or not"
  set_fact:
    unmet_requirements: "{{ unmet_requirements + [\"NVIDIA GPU card number is not matched: {{ computing_device_count }} specified but {{ nvidia_gpu_count.stdout|trim }} found\"] }}"
  changed_when: false
  check_mode: false
  environment: {}
  when:
    # if nvidia-smi doesn't work, we can skip this step
    - nvidia_smi.rc == 0
    - 'nvidia_gpu_count.stdout|trim|int != computing_device_count'
    - nvidia_gpu_count.rc == 0

- name: "Check default docker runtime"
  include_tasks: docker-runtime.yml