Re-use edpm-ansible driver installation code

[csibbitt]

• This is using the edpm-ansible role for an unintended purpose
• But it means we don't have to maintain a second code base

[csibbitt]

Testing

I tested this against our nova04delta cloud and it worked as expected. The VM was already provisioned before the test, so the output below is just for the driver installation and testing (--skip-tags rhoso).

$ cat inventory-local 
[control_node]
localhost ansible_connection=local

[target_nodes]
gpu-validation-0 ansible_host=192.168.25.222 ansible_user=cloud-user ansible_ssh_common_args='-i ~/.ssh/id_brq2_dataplane -o StrictHostKeyChecking=no -o ProxyCommand="ssh -W %h:%p -o StrictHostKeyChecking=no -i ~/.ssh/id_brq2_dataplane cloud-admin@<REDACTED>"'


$ cat vars-local.yaml 
---
gpu_validation_private_key_file: ~/.ssh/id_brq2_dataplane
gpu_validation_ca_cert_path: ./tls-ca-bundle.pem
gpu_validation_dns_server: 192.168.25.10
gpu_validation_floating_ip: 192.168.25.222
gpu_validation_ssh_proxy_command: '-o ProxyCommand="ssh -W %h:%p -o StrictHostKeyChecking=no -i ~/.ssh/id_brq2_dataplane cloud-admin@<REDACTED>"'
gpu_validation_pci_devices:
  10de:27b8: 1
gpu_validation_model_tests_enabled: false


$ JUNIT_OUTPUT_DIR=./ ansible-playbook --skip-tags rhoso -i inventory-local -e @vars-local.yaml main.yaml
[...]

TASK [Nvidia Setup (Use the edpm_accel_drivers role to install the NVIDIA driver in our VM)] ******************************************************************************************
included: osp.edpm.edpm_accel_drivers for gpu-validation-0
[...]

TASK [osp.edpm.edpm_accel_drivers : Add nvidia driver repo] *******************************************************************************************************************
changed: [gpu-validation-0]]
[...]

TASK [osp.edpm.edpm_accel_drivers : Enable NVIDIA driver DNF module for selected version stream] ******************************************************************************
changed: [gpu-validation-0]
[...]

TASK [osp.edpm.edpm_accel_drivers : Install the nvidia driver package] ********************************************************************************************************
changed: [gpu-validation-0]
[...]


TASK [gpu-validation : Run lspci command and save output] *********************************************************************************************************************
ok: [gpu-validation-0]

TASK [gpu-validation : Set found_nvidia to true if NVIDIA is found] ***********************************************************************************************************
ok: [gpu-validation-0]

TASK [gpu-validation : TEST[gpus] Check if GPUs in Passthrough mode are present in RHEL AI VM (lspci)] ************************************************************************
ok: [gpu-validation-0] => (item={'key': '10de:27b8', 'value': 1})
[...]

TASK [gpu-validation : TEST[nvidia] Check if GPUs in Passthrough mode are present in RHEL AI VM (nvidia-smi)] *****************************************************************
ok: [gpu-validation-0] => {
    "changed": false,
    "msg": "All assertions passed"
}
[...]

TASK [gpu-validation : TEST[CUDA] Run the CUDA Sanity Check] ******************************************************************************************************************
ok: [gpu-validation-0]
[...]

PLAY RECAP ********************************************************************************************************************************************************************
gpu-validation-0           : ok=31   changed=7    unreachable=0    failed=0    skipped=33   rescued=0    ignored=0   

[bogdando]

LGTM

[skovili]

I really like the idea of no second codebase to manage. and I see that the testing results are promising..so im ok to go ahead.

[MiguelCarpio]

LGTM