Run a Sample Workload#

After completing the deployment steps, you can verify your installation by running a sample GPU workload in a confidential container.

A pod manifest for a confidential container GPU workload requires that you specify the kata-qemu-nvidia-gpu-snp runtime class for SEV-SNP or kata-qemu-nvidia-gpu-tdx for TDX.

  1. Create a file, such as the following cuda-vectoradd-kata.yaml sample, specifying the kata-qemu-nvidia-gpu-snp runtime class:

    apiVersion: v1
kind: Pod
metadata:
  name: cuda-vectoradd-kata
  namespace: default
spec:
  runtimeClassName: kata-qemu-nvidia-gpu-snp
  restartPolicy: Never
  containers:
    - name: cuda-vectoradd
      image: "nvcr.io/nvidia/k8s/cuda-sample:vectoradd-cuda12.5.0-ubuntu22.04"
      resources:
        limits:
          nvidia.com/pgpu: "1"
          memory: 16Gi

    The following are Confidential Containers configurations in the sample manifest:

    • Set the runtime class to kata-qemu-nvidia-gpu-snp for SEV-SNP or kata-qemu-nvidia-gpu-tdx for TDX, depending on the node type where the workloads should run.

    • In the sample above, nvidia.com/pgpu is the default resource type for GPUs. If you are deploying on a heterogeneous cluster, you might want to update the default behavior by specifying the P_GPU_ALIAS environment variable for the sandbox device plugin. Refer to the Configuring the Sandbox Device Plugin to Use GPU or NVSwitch Specific Resource Types for more details.

    • If you have machines that support multi-GPU passthrough, refer to the Configuring Multi-GPU Passthrough page for a complete workload example and architecture-specific CC mode requirements.

  2. Create the pod:

    $ kubectl apply -f cuda-vectoradd-kata.yaml

    Example Output:

    pod/cuda-vectoradd-kata created

    Optional: Verify the pod is running.

    $ kubectl get pod cuda-vectoradd-kata

    Example Output:

    NAME                  READY   STATUS    RESTARTS   AGE
cuda-vectoradd-kata   1/1     Running   0          10s

  3. View the logs from the pod after the container starts:

    $ kubectl logs -n default cuda-vectoradd-kata

    Example Output:

    [Vector addition of 50000 elements]
Copy input data from the host memory to the CUDA device
CUDA kernel launch with 196 blocks of 256 threads
Copy output data from the CUDA device to the host memory
Test PASSED
Done

  4. Delete the pod:

    $ kubectl delete -f cuda-vectoradd-kata.yaml

Next Steps#

On this page