Auto-Create DRA Resources

[varunrsekar]

This PR adds new APIs to auto-create DRA resourceclaims.

At the bare minimum, the following input is needed:

spec:
  draResources:
  - claimSpec:
      devices:
      - name: gpu

This would generate the following resourceclaim spec:

spec:
  devices:
    requests:
    - exactly:
        allocationMode: ExactCount
        count: 1
        deviceClassName: gpu.nvidia.com
        selectors:
        - cel:
              expression: device.driver == "gpu.nvidia.com"
      name: gpu

For a more advanced auto-create experience:

spec:
  draResources:
  - claimSpec:
      isTemplate: true
      devices:
      - name: gpu
        deviceClassName: gpu.nvidia.com
        driverName: gpu.nvidia.com
        matchAttributes:
        - key: index
          op: NotEqual
          value:
            intValue: 1
        - key: driverVersion
          op: GreaterThanOrEqual
          value:
            versionValue: "550.127.8"
        - key: productName
          op: Equal
          value:
            stringValue: "NVIDIA A100-PCIE-40GB"
        matchCapacity:
        - key: memory
          value: 40Gi

This would generate the following resourceclaim template spec:

spec:
  devices:
    requests:
    - exactly:
        allocationMode: ExactCount
        count: 1
        deviceClassName: gpu.nvidia.com
        selectors:
        - cel:
              expression: device.driver == "gpu.nvidia.com"
          - cel:
              expression: device.attributes["gpu.nvidia.com"].index != 1
          - cel:
              expression: (device.attributes["gpu.nvidia.com"].driverVersion).compareTo(semver("550.127.8"))
                >= 0
          - cel:
              expression: device.attributes["gpu.nvidia.com"].productName == "NVIDIA
                A100-PCIE-40GB"
          - cel:
              expression: (device.capacity["gpu.nvidia.com"].memory).compareTo(quantity("40Gi"))
                == 0
      name: gpu

Notes:

1. The CEL expression device.driver == <spec.draResources[].claimSpec.devices[].driverName> is always added.
2. The label nim-operator.nvidia.com/dra-auto-generated="true" is always added.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[shivamerla]

Great to see this! thanks @varunrsekar