Files
iac-ansible-public/roles/nvidia/tasks/main.yml
2025-12-08 16:25:54 +00:00

94 lines
2.6 KiB
YAML

- name: Instalar as nvidia drivers
become: yes
become_user: fenix
shell: |
sudo apt install nvidia-driver-550
sudo apt install nvidia-utils-550
environment:
KUBECONFIG: /home/fenix/.kube/config
- name: Reboot remoto e continuar
become: yes
become_user: fenix
tasks:
- name: Reboot host e esperar voltar
ansible.builtin.reboot:
reboot_timeout: 600 # tempo máximo para o host voltar (segundos)
test_command: whoami # comando usado para validar que o host voltou
- name: Adicionar chave GPG da NVIDIA
become: yes
become_user: fenix
ansible.builtin.shell: |
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit.gpg
args:
creates: /usr/share/keyrings/nvidia-container-toolkit.gpg
- name: Adicionar repositório NVIDIA
become: yes
become_user: fenix
ansible.builtin.shell: |
curl -s -L https://nvidia.github.io/libnvidia-container/stable/ubuntu22.04/amd64/nvidia-container-toolkit.list | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit.gpg] https://#' | tee /etc/apt/sources.list.d/nvidia-container-toolkit.list
args:
creates: /etc/apt/sources.list.d/nvidia-container-toolkit.list
- name: Atualizar cache de pacotes
become: yes
become_user: fenix
ansible.builtin.apt:
update_cache: yes
- name: Instalar NVIDIA Container Toolkit
become: yes
become_user: fenix
ansible.builtin.apt:
name: nvidia-container-toolkit
state: present
- name: Configurar runtime Docker
become: yes
become_user: fenix
ansible.builtin.shell: |
nvidia-ctk runtime configure --runtime=docker
when: "'docker.io' in ansible_facts.packages"
- name: Reiniciar Docker
become: yes
become_user: fenix
ansible.builtin.systemd:
name: docker
state: restarted
when: "'docker.io' in ansible_facts.packages"
- name: Configurar runtime containerd
become: yes
become_user: fenix
ansible.builtin.shell: |
nvidia-ctk runtime configure --runtime=containerd
when: "'containerd' in ansible_facts.packages"
- name: Reiniciar containerd
become: yes
become_user: fenix
ansible.builtin.systemd:
name: containerd
state: restarted
when: "'containerd' in ansible_facts.packages"
- name: Validar GPU dentro de container
become: yes
become_user: fenix
ansible.builtin.shell: |
docker run --rm --runtime=nvidia --gpus all nvidia/cuda:12.2-base nvidia-smi
register: nvidia_smi_output
ignore_errors: true
- name: Mostrar resultado nvidia-smi
become: yes
become_user: fenix
ansible.builtin.debug:
var: nvidia_smi_output.stdout