Adds support for NVIDIA MIG configuration

Author: jovial

ansible/.gitignore

@@ -90,3 +90,5 @@ roles/*
 !roles/gateway/**
 !roles/alertmanager/
 !roles/alertmanager/**
+!roles/slurm_recompile/**
+!roles/slurm_recompile/**

ansible/extras.yml

@@ -48,6 +48,20 @@
         name: cuda
         tasks_from: "{{ 'runtime.yml' if appliances_mode == 'configure' else 'install.yml' }}"
 
+- name: Setup vGPU
+  hosts: vgpu
+  become: yes
+  gather_facts: yes
+  tags: vgpu
+  tasks:
+    - include_role:
+        name: stackhpc.linux.vgpu
+        tasks_from: "{{ 'configure.yml' if appliances_mode == 'configure' else 'install.yml' }}"
+  handlers:
+    - name: reboot
+      fail:
+        msg: Reboot handler for stackhpc.linux.vgpu role fired unexpectedly. This was supposed to be unreachable.
+
 - name: Persist hostkeys across rebuilds
   # Must be after filesystems.yml (for storage)
   # and before portal.yml (where OOD login node hostkeys are scanned)

ansible/fatimage.yml

@@ -250,6 +250,16 @@
         name: cloudalchemy.grafana
         tasks_from: install.yml
 
+- name: Add support for NVIDIA GPU auto detection to Slurm
+  hosts: cuda
+  become: yes
+  tasks:
+    - name: Recompile slurm
+      import_role:
+        name: slurm_recompile
+      vars:
+        slurm_recompile_with_nvml: "{{ groups.cuda | length > 0 }}"
+
 - name: Run post.yml hook
   vars:
     appliances_environment_root: "{{ lookup('env', 'APPLIANCES_ENVIRONMENT_ROOT') }}"

ansible/roles/compute_init/README.md

@@ -75,6 +75,7 @@ it also requires an image build with the role name added to the
 | extras.yml               | basic_users             | All functionality [6]           | No                  |
 | extras.yml               | eessi                   | All functionality [7]           | No                  |
 | extras.yml               | cuda                    | None required - use image build | Yes [8]             |
+| extras.yml               | vgpu                    | All functionality               | Yes                 |
 | extras.yml               | persist_hostkeys        | Not relevant for compute nodes  | n/a                 |
 | extras.yml               | compute_init (export)   | Not relevant for compute nodes  | n/a                 |
 | extras.yml               | k9s (install)           | Not relevant during boot        | n/a                 |

ansible/roles/compute_init/files/compute-init.yml

@@ -19,6 +19,8 @@
     enable_basic_users: "{{ os_metadata.meta.basic_users | default(false) | bool }}"
     enable_eessi: "{{ os_metadata.meta.eessi | default(false) | bool }}"
     enable_chrony: "{{ os_metadata.meta.chrony | default(false) | bool }}"
+    enable_vgpu: "{{ os_metadata.meta.vpgu | default(false) | bool }}"
+
 
     # TODO: "= role defaults" - could be moved to a vars_file: on play with similar precedence effects
     resolv_conf_nameservers: []
@@ -295,6 +297,12 @@
             cmd: "cvmfs_config setup"
       when: enable_eessi
 
+    - name: Configure VGPUs
+      include_role:
+        name: stackhpc.linux.vgpu
+        tasks_from: 'configure.yml'
+      when: enable_vgpu
+
     # NB: don't need conditional block on enable_compute as have already exited
     # if not the case
     - name: Write Munge key

ansible/validate.yml

@@ -88,3 +88,13 @@
     - import_role:
         name: lustre
         tasks_from: validate.yml
+
+- name: Validate vGPU configuration
+  hosts: vgpu
+  become: yes
+  gather_facts: yes
+  tags: vgpu
+  tasks:
+    - include_role:
+        name: stackhpc.linux.vgpu
+        tasks_from: validate.yml

environments/common/inventory/groups

@@ -112,6 +112,9 @@ freeipa_client
 [cuda]
 # Hosts to install NVIDIA CUDA on - see ansible/roles/cuda/README.md
 
+[vgpu]
+# Hosts where vGPU/MIG should be configured - see docs/mig.md
+
 [eessi]
 # Hosts on which EESSI stack should be configured
 

requirements.yml

@@ -4,7 +4,7 @@ roles:
     version: v25.3.2
     name: stackhpc.nfs
   - src: https://github.com/stackhpc/ansible-role-openhpc.git
-    version: v1.0.0
+    version: v1.1.0
     name: stackhpc.openhpc
   - src: https://github.com/stackhpc/ansible-node-exporter.git
     version: stackhpc
@@ -55,4 +55,6 @@ collections:
     version: 0.0.15
   - name: stackhpc.pulp
     version: 0.5.5
+  - name: stackhpc.linux
+    version: 1.4.0
 ...