beginnerEnablement ~15 min

Validate GPU availability in Kubernetes

Confirm that GPUs are discovered, enabled, and schedulable on your cluster.

Scenario

You've just enabled GPUs on a cluster and need to prove they're allocatable before onboarding workloads.

Prerequisites
  • ·kubectl access to a GPU cluster
  • ·The GPU Operator (or device plugin) installed
Tasks
0 / 3 tasks
  1. bash
    kubectl get nodes -o custom-columns=NAME:.metadata.name,GPU:.status.allocatable.'nvidia\.com/gpu'

    Expected · A number > 0 for each GPU node.

  2. bash
    kubectl -n gpu-operator get pods

    Expected · All pods Running / Completed.

  3. bash
    kubectl run cuda-check --rm -it --restart=Never \
  --image=nvidia/cuda:12.4.1-base-ubuntu22.04 \
  --limits=nvidia.com/gpu=1 -- nvidia-smi

    Expected · nvidia-smi prints the GPU table and the pod exits 0.

Success criteria
  • ✓Allocatable GPU count matches the hardware
  • ✓The test pod ran nvidia-smi successfully
Hints
  • ·If allocatable is 0, restart the device-plugin pod and check its logs.
  • ·If the host can't see the GPU, it's a driver/kernel issue — not Kubernetes.