From 28c9c0359d3d7db951f9b8e30357fc4c1c1fe249 Mon Sep 17 00:00:00 2001 From: Steve Brasier Date: Thu, 31 Jul 2025 13:32:06 +0000 Subject: [PATCH] make not rebuilding slurm for cuda (e.g. when using nvidia autodetection) possible --- ansible/fatimage.yml | 2 +- environments/common/inventory/groups | 3 +++ environments/common/layouts/everything | 4 ++++ 3 files changed, 8 insertions(+), 1 deletion(-) diff --git a/ansible/fatimage.yml b/ansible/fatimage.yml index ded3de31f..1c72416ee 100644 --- a/ansible/fatimage.yml +++ b/ansible/fatimage.yml @@ -262,7 +262,7 @@ name: grafana-dashboards - name: Add support for NVIDIA GPU auto detection to Slurm - hosts: cuda + hosts: slurm_recompile become: yes tasks: - name: Recompile slurm diff --git a/environments/common/inventory/groups b/environments/common/inventory/groups index cc4b57dce..0f9e204d5 100644 --- a/environments/common/inventory/groups +++ b/environments/common/inventory/groups @@ -118,6 +118,9 @@ freeipa_client [cuda] # Hosts to install NVIDIA CUDA on - see ansible/roles/cuda/README.md +[slurm_recompile] +# Hosts to recompile Slurm for - allows supporting Slurm autodetection method 'nvml' + [vgpu] # Hosts where vGPU/MIG should be configured - see docs/mig.md diff --git a/environments/common/layouts/everything b/environments/common/layouts/everything index 8c63247e6..0f4253758 100644 --- a/environments/common/layouts/everything +++ b/environments/common/layouts/everything @@ -65,6 +65,10 @@ cluster [cuda] # Hosts to install NVIDIA CUDA on - see ansible/roles/cuda/README.md +[slurm_recompile:children] +# Hosts to recompile Slurm for - allows supporting Slurm autodetection method 'nvml' +cuda + [eessi:children] # Hosts on which EESSI stack should be configured openhpc