Adds support for NVIDIA MIG configuration

Author: jovial

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: yeas
+  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 handlier 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/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.chrony | 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

@@ -83,3 +83,13 @@
     - 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

docs/mig.md

@@ -0,0 +1,17 @@
+# vGPU/MIG configuration
+
+Use variables from the [stackhpc.linux.vgpu](https://github.com/stackhpc/ansible-collection-linux/tree/main/roles/vgpu) role.
+
+For example in: `environments/<environment>/inventory/group_vars/all/vgpu`:
+
+```
+---
+vgpu_definitions:
+    - pci_address: "0000:17:00.0"
+      mig_devices:
+        "1g.10gb": 1
+        "2g.20gb": 3
+```
+
+The appliance will use the driver installed via the ``cuda`` role. Use ``lspci`` to determine the PCI
+addresses.

environments/common/inventory/group_vars/all/vgpu

@@ -0,0 +1,4 @@
+---
+
+# Nvidia driver is provided by cuda role.
+vgpu_nvidia_driver_install_enabled: false

environments/common/inventory/groups

@@ -112,6 +112,10 @@ freeipa_client
 [cuda]
 # Hosts to install NVIDIA CUDA on - see ansible/roles/cuda/README.md
 
+[vgpu]
+# FIXME: Update once PR merged
+# Hosts where vGPU/MIG should be configured - see https://github.com/stackhpc/ansible-collection-linux/pull/43/files#diff-74e43d9a34244aa54721f4dbd12a029baa87957afd762b88c2677aa75414f514R75
+
 [eessi]
 # Hosts on which EESSI stack should be configured
 

requirements.yml

@@ -55,4 +55,7 @@ collections:
     version: 0.0.15
   - name: stackhpc.pulp
     version: 0.5.5
+  - name: https://github.com/stackhpc/ansible-collection-linux
+    type: git
+    version: feature/mig-only
 ...