Skip to content

Configure GPU (vGPU) for Virtual Machines

This page will explain the prerequisites for configuring GPU when creating a virtual machine.

The key to configuring GPU for virtual machines is to configure the GPU Operator to deploy different software components on the worker nodes, depending on the GPU workload configuration. Here are three example nodes:

  • The controller-node-1 node is configured to run containers.
  • The work-node-1 node is configured to run virtual machines with GPU passthrough.
  • The work-node-2 node is configured to run virtual machines with virtual vGPU.

Assumptions, Limitations, and Dependencies

The worker nodes can run GPU-accelerated containers, virtual machines with GPU passthrough, or virtual machines with vGPU. However, a combination of any of these is not supported.

  1. The worker nodes can run GPU-accelerated containers, virtual machines with GPU passthrough, or virtual machines with vGPU individually, but not in any combination.
  2. The cluster administrator or developer needs to have prior knowledge of the cluster and correctly label the nodes to indicate the type of GPU workload they will run.
  3. The worker node that runs a GPU-accelerated virtual machine with GPU passthrough or vGPU is assumed to be a bare metal machine. If the worker node is a virtual machine, the GPU passthrough feature needs to be enabled on the virtual machine platform. Please consult your virtual machine platform provider for guidance.
  4. Nvidia MIG is not supported for vGPU.
  5. The GPU Operator does not automatically install GPU drivers in the virtual machine.

Enable IOMMU

To enable GPU passthrough, the cluster nodes need to have IOMMU enabled. Please refer to How to Enable IOMMU. If your cluster is running on a virtual machine, please consult your virtual machine platform provider.

Build vGPU Manager Image

Note: This step is only required when using NVIDIA vGPU. If you plan to use GPU passthrough only, skip this section.

Follow these steps to build the vGPU Manager image and push it to the container registry:

  1. Download the vGPU software from the NVIDIA Licensing Portal.

    • Log in to the NVIDIA Licensing Portal and go to the Software Downloads page.
    • The NVIDIA vGPU software is located in the Driver downloads tab on the Software Downloads page.
    • Select VGPU + Linux in the filter criteria and click Download to get the Linux KVM package. Unzip the downloaded file (NVIDIA-Linux-x86_64-<version>-vgpu-kvm.run).

  2. Open a terminal and clone the container-images/driver repository.

    git clone https://gitlab.com/nvidia/container-images/driver cd driver

  3. Switch to the vgpu-manager directory corresponding to your operating system.

    cd vgpu-manager/<your-os>

  4. Copy the .run file extracted in step 1 to the current directory.

    cp <local-driver-download-directory>/*-vgpu-kvm.run ./

  5. Set the environment variables.

    • PRIVATE_REGISTRY: The name of the private registry to store the driver image.
    • VERSION: The version of the NVIDIA vGPU Manager, downloaded from the NVIDIA Software Portal.
    • OS_TAG: Must match the operating system version of the cluster nodes.
    • CUDA_VERSION: The base CUDA image version used to build the driver image.
    export PRIVATE_REGISTRY=my/private/registry VERSION=510.73.06 OS_TAG=ubuntu22.04 CUDA_VERSION=12.2.0

  6. Build the NVIDIA vGPU Manager Image.

    docker build \
  --build-arg DRIVER_VERSION=${VERSION} \
  --build-arg CUDA_VERSION=${CUDA_VERSION} \
  -t ${PRIVATE_REGISTRY}``/vgpu-manager``:${VERSION}-${OS_TAG} .

  7. Push the NVIDIA vGPU Manager image to your container registry.

    docker push ${PRIVATE_REGISTRY}/vgpu-manager:${VERSION}-${OS_TAG}

Label the Cluster Nodes

Go to Container Management, select your worker cluster, click Node Management, and then click Modify Labels in the action bar to add labels to the nodes. Each node can only have one label.

You can assign the following values to the labels: container, vm-passthrough, and vm-vgpu.

Install Nvidia Operator

  1. Go to Container Management, select your worker cluster, click Helm Apps -> Helm Chart, and choose and install gpu-operator. Modify the relevant fields in the yaml.

    gpu-operator.sandboxWorkloads.enabled=true
gpu-operator.vgpuManager.enabled=true
gpu-operator.vgpuManager.repository=<your-register-url>      # (1)!
gpu-operator.vgpuManager.image=vgpu-manager
gpu-operator.vgpuManager.version=<your-vgpu-manager-version> # (2)!
gpu-operator.vgpuDeviceManager.enabled=true
    1. The container registry address from the "Build vGPU Manager Image" step.
    2. The VERSION from the "Build vGPU Manager Image" step.

  2. Wait for the installation to succeed, as shown in the following image:

Install virtnest-agent and Configure CR

  1. Install virtnest-agent, refer to Install virtnest-agent.

  2. Add vGPU and GPU passthrough to the Virtnest Kubevirt CR. The following example shows the relevant yaml after adding vGPU and GPU passthrough:

    spec:
  configuration:
    developerConfiguration:
      featureGates:
      - GPU
      - DisableMDEVConfiguration
    permittedHostDevices: # (1)!
      mediatedDevices:            # (2)!
      - mdevNameSelector: GRID P4-1Q
        resourceName: nvidia.com /GRID_P4-1Q
      pciHostDevices:             # (3)!
      - externalResourceProvider:  true
        pciVendorSelector: 10DE:1BB3
        resourceName: nvidia.com /GP104GL_TESLA_P4
    1. The following information needs to be filled in
    2. vGPU
    3. GPU passthrough

  3. In the kubevirt CR yaml, permittedHostDevices is used to import VM devices. For vGPU, mediatedDevices needs to be added, with the following structure:

    mediatedDevices:          
- mdevNameSelector: GRID P4-1Q          # (1)!
  resourceName: nvidia.com/GRID_P4-1Q   # (2)!
    1. Device name
    2. vGPU information registered by the GPU Operator on the node

  4. For GPU passthrough, pciHostDevices needs to be added under permittedHostDevices, with the following structure:

    pciHostDevices:           
- externalResourceProvider: true            # (1)!
  pciVendorSelector: 10DE:1BB3              # (2)!
  resourceName: nvidia.com/GP104GL_TESLA_P4 # (3)!
    1. Do not change this by default
    2. Vendor ID of the current PCI device
    3. GPU information registered by the GPU Operator on the node

  5. Example of obtaining vGPU information (only applicable to vGPU): View the node information on the node marked as nvidia.com/gpu.workload.config=vm-gpu, such as work-node-2, in the Capacity section, nvidia.com/GRID_P4-1Q: 8 indicates the available vGPU:

    kubectl describe node work-node-2

    Capacity:
  cpu:                                 64
  devices.kubevirt.io/kvm:             1k
  devices.kubevirt.io/tun:             1k
  devices.kubevirt.io/vhost-net:       1k
  ephemeral-storage:                   102626232Ki
  hugepages-1Gi:                       0
  hugepages-2Mi:                       0
  memory:                              264010840Ki
  nvidia.com/GRID_P4-1Q :              8
  pods:                                110
Allocatable:
  cpu:                                  64
  devices.kubevirt.io/kvm:              1k
  devices.kubevirt.io/tun:              1k
  devices.kubevirt.io/vhost-net:        1k
  ephemeral-storage:                    94580335255
  hugepages-1Gi:                        0
  hugepages-2Mi:                        0
  memory:                               263908440Ki
  nvidia.com/GRID_P4-1Q:                8
  pods:                                 110

    In this case, the mdevNameSelector should be "GRID P4-1Q" and the resourceName should be "GRID_P4-1Q".

  6. Get GPU passthrough information: On the node marked as nvidia.com/gpu.workload.config=vm-passthrough (work-node-1 in this example), view the node information. In the Capacity section, nvidia.com/GP104GL_TESLA_P4: 2 represents the available vGPU:

    kubectl describe node work-node-1

    Capacity:
  cpu:                            64
  devices.kubevirt.io/kvm:        1k
  devices.kubevirt.io/tun:        1k
  devices.kubevirt.io/vhost-net:  1k
  ephemeral-storage:              102626232Ki
  hugepages-1Gi:                  0
  hugepages-2Mi:                  0
  memory:                         264010840Ki
  nvidia.com/GP104GL_TESLA_P4:    2
  pods:                           110
Allocatable:
  cpu:                            64
  devices.kubevirt.io/kvm:        1k
  devices.kubevirt.io/tun:        1k
  devices.kubevirt.io/vhost-net:  1k
  ephemeral-storage:              94580335255
  hugepages-1Gi:                  0
  hugepages-2Mi:                  0
  memory:                         263908440Ki
  nvidia.com/GP104GL_TESLA_P4:    2
  pods:                           110

    The resourceName should be "GRID_P4-1Q". How to obtain the pciVendorSelector? SSH into the target node work-node-1 and use the lspci -nnk -d 10de: command to obtain the Nvidia GPU PCI information, as shown below: The red box indicates the pciVendorSelector information.

  7. Edit the kubevirt CR note: If there are multiple GPUs of the same model, you only need to write one in the CR, there is no need to list every GPU.

    kubectl -n virtnest-system edit kubevirt kubevirt

    spec:
  configuration:
    developerConfiguration:
      featureGates:
      - GPU
      - DisableMDEVConfiguration
    permittedHostDevices: # (1)!
      mediatedDevices:                    # (2)!
      - mdevNameSelector: GRID P4-1Q
        resourceName: nvidia.com/GRID_P4-1Q
      pciHostDevices:                       # (3)!
      - externalResourceProvider: true
        pciVendorSelector: 10DE:1BB3
        resourceName: nvidia.com/GP104GL_TESLA_P4

    1. The following information needs to be filled in
    2. vGPU
    3. GPU passthrough; in the example above, there are two GPUs for TEESLA P4, so only one needs to be registered here

Create VM Using YAML and Enable GPU Acceleration

The only difference from a regular virtual machine is adding the gpu-related information in the devices section.

Click to view the complete YAML 
apiVersion: kubevirt.io/v1
kind: VirtualMachine
metadata:
  name: testvm-gpu1
  namespace: default
spec:
  dataVolumeTemplates:
  - metadata:
      creationTimestamp: null
      name: systemdisk-testvm-gpu1
      namespace: default
    spec:
      pvc:
        accessModes:
        - ReadWriteOnce
        resources:
          requests:
            storage: 10Gi
        storageClassName: www
      source:
        registry:
          url: docker://release-ci.daocloud.io/virtnest/system-images/debian-12-x86_64:v1
runStrategy: Manual
template:
    metadata:
      creationTimestamp: null
    spec:
      domain:
        cpu:
          cores: 1
          sockets: 1
          threads: 1
        devices:
          disks:
          - bootOrder: 1
            disk:
              bus: virtio
            name: systemdisk-testvm-gpu1
          - disk:
              bus: virtio
            name: cloudinitdisk
        gpus:
        - deviceName: nvidia.com/GP104GL_TESLA_P4
            name: gpu-0-0
        - deviceName: nvidia.com/GP104GL_TESLA_P4
          name: gpu-0-1
        interfaces:
        - masquerade: {}
          name: default
      machine:
        type: q35
      resources:
        requests:
          memory: 2Gi
    networks:
    - name: default
      pod: {}
    volumes:
    - dataVolume:
        name: systemdisk-testvm-gpu1
      name: systemdisk-testvm-gpu1
    - cloudInitNoCloud:
        userDataBase64: I2Nsb3VkLWNvbmZpZwpzc2hfcHdhdXRoOiB0cnVlCmRpc2FibGVfcm9vdDogZmFsc2UKY2hwYXNzd2Q6IHsibGlzdCI6ICJyb290OmRhbmdlcm91cyIsIGV4cGlyZTogRmFsc2V9CgoKcnVuY21kOgogIC0gc2VkIC1pICIvI1w/UGVybWl0Um9vdExvZ2luL3MvXi4qJC9QZXJtaXRSb290TG9naW4geWVzL2ciIC9ldGMvc3NoL3NzaGRfY29uZmlnCiAgLSBzeXN0ZW1jdGwgcmVzdGFydCBzc2guc2VydmljZQ==
        name: cloudinitdisk

Comments