From fa78672bf8267abe189a842d6791464d6dec1903 Mon Sep 17 00:00:00 2001 From: Claudia Watson Date: Tue, 28 Oct 2025 15:27:31 +0000 Subject: [PATCH] Adding tasks to essi/configure.yml to make eessi configure gpu node automatically --- .../roles/compute_init/files/compute-init.yml | 16 ++++++++++++++++ ansible/roles/eessi/tasks/configure.yml | 17 +++++++++++++++++ 2 files changed, 33 insertions(+) diff --git a/ansible/roles/compute_init/files/compute-init.yml b/ansible/roles/compute_init/files/compute-init.yml index 81dedf8fb..0a21303a6 100644 --- a/ansible/roles/compute_init/files/compute-init.yml +++ b/ansible/roles/compute_init/files/compute-init.yml @@ -294,6 +294,22 @@ ansible.builtin.command: cmd: "cvmfs_config setup" + # configure gpus + - name: Check for NVIDIA driver + ansible.builtin.stat: + path: /dev/nvidia0 + register: nvidia_driver + + - name: Set fact if NVIDIA driver is present + ansible.builtin.set_fact: + has_nvidia_driver: "{{ nvidia_driver.stat.exists | default(false) }}" + + - name: Expose GPU drivers + ansible.builtin.shell: | + source /cvmfs/software.eessi.io/versions/2023.06/init/bash + /cvmfs/software.eessi.io/versions/2023.06/scripts/gpu_support/nvidia/link_nvidia_host_libraries.sh + when: has_nvidia_driver + - name: Configure VGPUs ansible.builtin.include_role: name: stackhpc.linux.vgpu diff --git a/ansible/roles/eessi/tasks/configure.yml b/ansible/roles/eessi/tasks/configure.yml index 2c765d20c..23030c3e5 100644 --- a/ansible/roles/eessi/tasks/configure.yml +++ b/ansible/roles/eessi/tasks/configure.yml @@ -15,3 +15,20 @@ - name: Ensure CVMFS config is setup # noqa: no-changed-when ansible.builtin.command: cmd: "cvmfs_config setup" + +# configure gpus +- name: Check for NVIDIA driver + ansible.builtin.stat: + path: /dev/nvidia0 + register: nvidia_driver + +- name: Set fact if NVIDIA driver is present + ansible.builtin.set_fact: + has_nvidia_driver: "{{ nvidia_driver.stat.exists | default(false) }}" + +- name: Expose GPU drivers + ansible.builtin.shell: | + source /cvmfs/software.eessi.io/versions/2023.06/init/bash + /cvmfs/software.eessi.io/versions/2023.06/scripts/gpu_support/nvidia/link_nvidia_host_libraries.sh + when: has_nvidia_driver + changed_when: true