Files
cluster-knaster/platform/proxmox/ansible/roles/gpu-node/tasks/main.yml
T

85 lines
2.6 KiB
YAML

- name: Treiber-Branch bestimmen
ansible.builtin.set_fact:
_branch: "{{ nvidia_driver_branch[driver_generation] }}"
- name: NVIDIA-Treiber (Server-Variante) installieren
ansible.builtin.apt:
name:
- "nvidia-driver-{{ _branch }}-server"
- "nvidia-utils-{{ _branch }}-server"
state: present
update_cache: true
register: driver
- name: Treiber gegen ungewollte Updates sperren
ansible.builtin.dpkg_selections:
name: "{{ item }}"
selection: hold
loop:
- "nvidia-driver-{{ _branch }}-server"
- "nvidia-utils-{{ _branch }}-server"
- name: NVIDIA-Container-Toolkit-Repository
ansible.builtin.shell: |
set -euo pipefail
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg
curl -fsSL https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list \
| sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' \
> /etc/apt/sources.list.d/nvidia-container-toolkit.list
args:
creates: /etc/apt/sources.list.d/nvidia-container-toolkit.list
executable: /bin/bash
- name: Container-Toolkit installieren und für Docker konfigurieren
ansible.builtin.apt:
name: nvidia-container-toolkit
state: present
update_cache: true
notify: configure nvidia runtime
- name: Modell-Disk (scsi1) formatieren, falls leer (formatiert nie eine vorhandene Partition um)
community.general.filesystem:
dev: /dev/sdb
fstype: ext4
opts: -L models
when: ansible_devices.sdb is defined
- name: Modell-Disk einhängen
ansible.posix.mount:
path: /srv/models
src: LABEL=models
fstype: ext4
state: mounted
when: ansible_devices.sdb is defined
- name: Textfile-Verzeichnis für node_exporter
ansible.builtin.file:
path: "{{ nvidia_textfile_dir }}"
state: directory
mode: "0755"
- name: Xid-Exporter-Skript
ansible.builtin.copy:
src: "{{ playbook_dir }}/../../../../deploy/scripts/gpu-xid-textfile.sh"
dest: /usr/local/bin/gpu-xid-textfile.sh
mode: "0755"
- name: systemd-Service und -Timer für Xid-Exporter
ansible.builtin.template:
src: "{{ item }}.j2"
dest: "/etc/systemd/system/{{ item }}"
mode: "0644"
loop: [gpu-xid-textfile.service, gpu-xid-textfile.timer]
notify: reload systemd
- name: Timer aktivieren
ansible.builtin.systemd:
name: gpu-xid-textfile.timer
enabled: true
state: started
daemon_reload: true
- name: Neustart nach Treiberinstallation
ansible.builtin.reboot:
when: driver.changed