From de76ee42e552da1fefb9467ea2e0835642493117 Mon Sep 17 00:00:00 2001 From: Claudia Watson Date: Thu, 23 Oct 2025 13:28:07 +0100 Subject: [PATCH 1/2] Add option to install nvidia-fabricmanager --- ansible/roles/cuda/defaults/main.yml | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/ansible/roles/cuda/defaults/main.yml b/ansible/roles/cuda/defaults/main.yml index 692301d23..5790953e3 100644 --- a/ansible/roles/cuda/defaults/main.yml +++ b/ansible/roles/cuda/defaults/main.yml @@ -2,13 +2,17 @@ # yamllint disable-line rule:line-length cuda_repo_url: "https://developer.download.nvidia.com/compute/cuda/repos/rhel{{ ansible_distribution_major_version }}/{{ ansible_architecture }}/cuda-rhel{{ ansible_distribution_major_version }}.repo" cuda_nvidia_driver_stream: '580-open' -cuda_nvidia_driver_pkg: "nvidia-open-3:580.82.07-1.el{{ ansible_distribution_major_version }}" +cuda_nvidia_driver_version: '580.82.07-1' +cuda_nvidia_driver_pkg: "nvidia-open-3:{{ cuda_nvidia_driver_version }}.el{{ ansible_distribution_major_version }}" cuda_package_version: '13.0.1-1' cuda_version_short: "{{ (cuda_package_version | split('.'))[0:2] | join('.') }}" # major.minor -cuda_packages: +cuda_packages_default: - "cuda-toolkit-{{ cuda_package_version }}" - nvidia-gds - cmake +cuda_packages_fabricmanager: + - "nvidia-fabricmanager-{{ cuda_nvidia_driver_version }}" +cuda_packages: "{{ cuda_packages_default + ( cuda_packages_fabricmanager if cuda_install_nvidiafabricmanger | bool else [] ) }}" cuda_samples_release_url: "https://github.com/NVIDIA/cuda-samples/archive/refs/tags/v{{ cuda_version_short }}.tar.gz" cuda_samples_path: "/var/lib/{{ ansible_user }}/cuda_samples" cuda_samples_programs: @@ -16,3 +20,5 @@ cuda_samples_programs: - bandwidthTest # cuda_devices: # discovered from deviceQuery run cuda_persistenced_state: started +# option whether or not to install nvidia-fabricmanager default false +cuda_install_nvidiafabricmanger: false From 7603066b71de9fb5a0a5ec99c10ed7c6d1b728ea Mon Sep 17 00:00:00 2001 From: Claudia Watson Date: Fri, 24 Oct 2025 15:24:00 +0100 Subject: [PATCH 2/2] Remove commented cuda_install_nvidiafabricmanger option Remove commented option for installing nvidia-fabricmanager. --- ansible/roles/cuda/defaults/main.yml | 1 - 1 file changed, 1 deletion(-) diff --git a/ansible/roles/cuda/defaults/main.yml b/ansible/roles/cuda/defaults/main.yml index 5790953e3..19720309c 100644 --- a/ansible/roles/cuda/defaults/main.yml +++ b/ansible/roles/cuda/defaults/main.yml @@ -20,5 +20,4 @@ cuda_samples_programs: - bandwidthTest # cuda_devices: # discovered from deviceQuery run cuda_persistenced_state: started -# option whether or not to install nvidia-fabricmanager default false cuda_install_nvidiafabricmanger: false