Quick Start Guide

Quick Start Guide#

Getting up and running with the AMD GPU Operator and Device Metrics Exporter on Kubernets is quick and easy. Below is a short guide on how to get started using the helm installation method on a standard Kubernetes install. Note that more detailed instructions on the different installation methods can be found on this site:
GPU Operator Kubernets Helm Install
GPU Operator Red Hat OpenShift Install

Installing the GPU Operator#

  1. The GPU Operator uses cert-manager to manage certificates for MTLS communication between services. If you haven’t already installed cert-manager as a prerequisite on your Kubernetes cluster, you’ll need to install it as follows:

    # Add and update the cert-manager repository
helm repo add jetstack https://charts.jetstack.io --force-update

# Install cert-manager
helm install cert-manager jetstack/cert-manager \
  --namespace cert-manager \
  --create-namespace \
  --version v1.15.1 \
  --set crds.enabled=true

  2. Once cert-manager is installed, you’re just a few commands away from installing the GPU Operating and having a fully managed GPU infrastructure:

    # Add the Helm repository
helm repo add rocm https://rocm.github.io/gpu-operator
helm repo update

# Install the GPU Operator
helm install amd-gpu-operator rocm/gpu-operator-charts \
  --namespace kube-amd-gpu --create-namespace

    Optionally, you can use the Helm --version parameter to install a specific version of the GPU Operator (v1.2.0 in this case) as follows:

    helm install amd-gpu-operator rocm/gpu-operator-charts \
  --namespace kube-amd-gpu --create-namespace \
  --version=1.2.0

  3. You should now see the GPU Operator component pods starting up in the namespace you specified above, kube-amd-gpu. You will also notice that the gpu-operator-charts-controller-manager, kmm-controller and kmm-webhook-server pods are in a pending state. This is because you need to label a node in your cluster as the control-plane node for those pods to run on:

    # Label the control-plane node
kubectl label nodes <node-name> node-role.kubernetes.io/control-plane=

  4. To deploy the Device Plugin, Node Labeller and Metrics exporter to your cluster you need to create a new DeviceConfig custom resource. For a full list of configurable options refer to the Full Reference Config documenattion. An example DeviceConfig is supplied in the ROCm/gpu-operator repository which can be used to get going:

    # Apply the example DeviceConfig to enable the Device Plugin, Node Labeller and Metrics Exporter plugins
kubectl apply -f https://raw.githubusercontent.com/ROCm/gpu-operator/refs/heads/main/example/deviceconfig_example.yaml

    Note

    If you are using a previous version of the GPU Operator you need to apply the deviceconfig_example.yaml file from that specific branch (v1.1.0 in this example) by doing the following:

    kubectl apply -f https://raw.githubusercontent.com/ROCm/gpu-operator/refs/heads/release-v1.1.0/example/deviceconfig_example.yaml

That's it! The GPU Operator components should now all be running. You can verify this by checking the namespace where the gpu-operator components are installed (default: `kube-amd-gpu`): 
kubectl get pods -n kube-amd-gpu

Creating a GPU-enabled Pod#

To create a pod that uses a GPU, specify the GPU resource in your pod specification:

apiVersion: v1
kind: Pod
metadata:
  name: gpu-pod
spec:
  containers:
    - name: gpu-container
      image: rocm/pytorch:latest
      resources:
        limits:
          amd.com/gpu: 1 # requesting 1 GPU

Save this YAML to a file (e.g., gpu-pod.yaml) and create the pod:

kubectl apply -f gpu-pod.yaml

Checking GPU Status#

To check the status of GPUs in your cluster:

kubectl get nodes -o custom-columns=NAME:.metadata.name,GPUs:.status.capacity.'amd\.com/gpu'

Using amd-smi#

To run amd-smi in a pod:

  • Create a YAML file named amd-smi.yaml:

apiVersion: v1
kind: Pod
metadata:
 name: amd-smi
spec:
 containers:
 - image: docker.io/rocm/pytorch:latest
   name: amd-smi
   command: ["/bin/bash"]
   args: ["-c","amd-smi version && amd-smi monitor -ptum"]
   resources:
    limits:
      amd.com/gpu: 1
    requests:
      amd.com/gpu: 1
 restartPolicy: Never

  • Create the pod:

kubectl create -f amd-smi.yaml

  • Check the logs and verify the output amd-smi reflects the expected ROCm version and GPU presence:

kubectl logs amd-smi
AMDSMI Tool: 24.6.2+2b02a07 | AMDSMI Library version: 24.6.2.0 | ROCm version: 6.2.2
GPU  POWER  GPU_TEMP  MEM_TEMP  GFX_UTIL  GFX_CLOCK  MEM_UTIL  MEM_CLOCK
  0  126 W     40 °C     32 °C       1 %    182 MHz       0 %    900 MHz

Using rocminfo#

To run rocminfo in a pod:

  • Create a YAML file named rocminfo.yaml:

apiVersion: v1
kind: Pod
metadata:
  name: rocminfo
spec:
  containers:
  - image: rocm/pytorch:latest
    name: rocminfo
    command: ["/bin/sh","-c"]
    args: ["rocminfo"]
    resources:
      limits:
        amd.com/gpu: 1
  restartPolicy: Never

  • Create the pod:

kubectl create -f rocminfo.yaml

  • Check the logs and verify the output:

kubectl logs rocminfo

Configuring GPU Resources#

Configuration parameters are documented in the Custom Resource Installation Guide

Contents