Support automatic GRES configuration for NVIDIA GPUs (#820)

* wip - bump openhpc role for testing

* remove GresTypes from MIG docs

* enable nvml autoconfiguration for CaaS

* fix linter problems

* bump openhpc to release w/ auto gres support

Author: sjpb

docs/mig.md

@@ -214,10 +214,6 @@ openhpc_nodegroups:
       - conf: "gpu:nvidia_h100_80gb_hbm3:2"
       - conf: "gpu:nvidia_h100_80gb_hbm3_4g.40gb:2"
       - conf: "gpu:nvidia_h100_80gb_hbm3_1g.10gb:6"
-
-openhpc_config:
-  GresTypes:
-    - gpu
 ```
 
 Making sure the types (the identifier after `gpu:`) match those collected with `slurmd -G`. Substrings

environments/.caas/inventory/group_vars/all/openhpc.yml

@@ -4,3 +4,6 @@ openhpc_cluster_name: "{{ cluster_name }}"
 # Provision a single "standard" compute nodegroup using the supplied
 # node count and flavor
 openhpc_nodegroups: "{{ hostvars[groups['openstack'][0]]['openhpc_nodegroups'] }}"
+
+# Enable autoconfiguration of NVIDIA GPUs, if using a suitable (`cuda`) image:
+openhpc_gres_autodetect: nvml

requirements.yml

@@ -4,7 +4,7 @@ roles:
     version: v25.3.2
     name: stackhpc.nfs
   - src: https://github.com/stackhpc/ansible-role-openhpc.git
-    version: v1.4.1
+    version: v1.5.0
     name: stackhpc.openhpc
   - src: https://github.com/stackhpc/ansible-node-exporter.git
     version: stackhpc