Enhancement of nvidia-device-plugin-daemonset in gpu-cluster.md

[JoeyC-Dev]

Proposed change: Improve the Nvidia device plugin deamonset for better deployment.

Supporting point:

1. Currently, if the user is using Spot node, the current version of deamonset will FAIL to deploy due to missing tolerations to relative taint.
2. Also, the current daemonset will also try to deploy the Pod to AMD GPU / non-GPU node due to missing nodeSelector.
3. As of now, the latest version of plugin is v0.17.0. Update image version relatively: https://github.com/NVIDIA/k8s-device-plugin/releases
4. The use of label/taint is following the proposed/supported list:
   • https://learn.microsoft.com/en-us/azure/aks/use-labels#reserved-system-labels
   • https://learn.microsoft.com/en-us/azure/aks/spot-node-pool#:~:text=kubernetes.azure.com/scalesetpriority=spot:NoSchedule
5. Weirdly, currently in the document is asking for creating namespace gpu-operator, but there is no actual use in the daemonset. Consider this is a manual set-up, I kept it as in namespace gpu-operator and change the yaml relatively.
   

   azure-aks-docs/articles/aks/gpu-cluster.md

   Lines 128 to 132 in 130a484

   	1. Create a namespace using the [`kubectl create namespace`][kubectl-create] command.
   	```bash
   	kubectl create namespace gpu-operator
   	```

Having checked, the propsed yaml is working, and it will deploy the ds on Nvidia GPU node.

Successful deployment:

kubectl logs nvidia-device-plugin-daemonset-7fzlq -n gpu-operator
I0114 09:03:01.133253       1 main.go:235] "Starting NVIDIA Device Plugin" version=<
        d475b2cf
        commit: d475b2cfcf12b983a4975d4fc59d91af432cf28e
 >
I0114 09:03:01.133332       1 main.go:238] Starting FS watcher for /var/lib/kubelet/device-plugins
I0114 09:03:01.133367       1 main.go:245] Starting OS watcher.
I0114 09:03:01.133653       1 main.go:260] Starting Plugins.
I0114 09:03:01.133684       1 main.go:317] Loading configuration.
I0114 09:03:01.134351       1 main.go:342] Updating config with default resource matching patterns.
I0114 09:03:01.134496       1 main.go:353] 
Running with config:
{
  "version": "v1",
  "flags": {
    "migStrategy": "none",
    "failOnInitError": false,
    "mpsRoot": "",
    "nvidiaDriverRoot": "/",
    "nvidiaDevRoot": "/",
    "gdsEnabled": false,
    "mofedEnabled": false,
    "useNodeFeatureAPI": null,
    "deviceDiscoveryStrategy": "auto",
    "plugin": {
      "passDeviceSpecs": false,
      "deviceListStrategy": [
        "envvar"
      ],
      "deviceIDStrategy": "uuid",
      "cdiAnnotationPrefix": "cdi.k8s.io/",
      "nvidiaCTKPath": "/usr/bin/nvidia-ctk",
      "containerDriverRoot": "/driver-root"
    }
  },
  "resources": {
    "gpus": [
      {
        "pattern": "*",
        "name": "nvidia.com/gpu"
      }
    ]
  },
  "sharing": {
    "timeSlicing": {}
  },
  "imex": {}
}
I0114 09:03:01.134503       1 main.go:356] Retrieving plugins.
I0114 09:03:01.162986       1 server.go:195] Starting GRPC server for 'nvidia.com/gpu'
I0114 09:03:01.164044       1 server.go:139] Starting to serve 'nvidia.com/gpu' on /var/lib/kubelet/device-plugins/nvidia-gpu.sock
I0114 09:03:01.166773       1 server.go:146] Registered device plugin for 'nvidia.com/gpu' with Kubelet

[prmerger-automator]

@JoeyC-Dev : Thanks for your contribution! The author(s) have been notified to review your proposed change.

[learn-build-service-prod]

Learn Build status updates of commit 07ab281:

✅ Validation status: passed

File	Status	Preview URL	Details
articles/aks/gpu-cluster.md	✅Succeeded

For more details, please refer to the build report.

For any questions, please:

• Try searching the learn.microsoft.com contributor guides
• Post your question in the Learn support channel

[ttorble]

@schaffererin

Can you review the proposed changes?

IMPORTANT: When the changes are ready for publication, adding a #sign-off comment is the best way to signal that the PR is ready for the review team to merge.

#label:"aq-pr-triaged"
@MicrosoftDocs/public-repo-pr-review-team

[github-actions]

This pull request has been inactive for at least 14 days. If you are finished with your changes, don't forget to sign off. See the contributor guide for instructions.
Get Help
Docs Support Teams Channel
Resolve Merge Conflict

[v-dirichards]

@schaffererin This PR hasn’t had any updates for a while. If it's ready for review, could you sign off? Or should it be closed?

#label:"aq-pr-triaged","aq-followed-up"

[github-actions]

This pull request has been inactive for at least 14 days. If you are finished with your changes, don't forget to sign off. See the contributor guide for instructions.
Get Help
Docs Support Teams Channel
Resolve Merge Conflict

[JoeyC-Dev]

This PR is critical for users who aren't expert on Kubernetes as they don't know why their driver cannot be installed.
Can anyone kindly approve this PR?

@MicrosoftDocs/public-repo-pr-review-team

[Court72]

I sent an email to the content owner today.