NVKind Cluster Deployment with GPUs

Overview

This section targets cases where you have a host machine (local or remote) that has access to NVIDIA GPUs and leverage nvkind to create and manage kind kubernetes clusters with access to GPUs.

The automated shell script will deploy dependencies (Minio, Keycloak and Postgresql), before generating the deployment configuration for the studio and then deploying the main studio services + pipelines.

---

Prerequisites

Required Tools

Tool	Version	Installation
nvkind	Latest	Install Guide
Helm	v3.19	Install Helm
OpenShift CLI	Latest	Install oc
kubectl	Latest	Bundled with oc
jq	Latest	Install jq
yq	Latest	Install yq

Helm Version

v3.19 required - incompatible with v4

Deployment Steps

Step 1: Configure nvkind Prerequisites

Follow the nvkind prerequisites for your OS.

Verify GPU detection:

$ nvidia-smi -L

Expected output:

# GPU 0: NVIDIA L4 (UUID: GPU-3e71c48d-90c0-f46f-195b-4150320f9910)

$ docker run --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=all ubuntu:20.04 nvidia-smi -L

Expected output:

# GPU 0: NVIDIA L4 (UUID: GPU-3e71c48d-90c0-f46f-195b-4150320f9910)

---

Step 2: Run nvkind Setup

Complete the setup steps from nvkind documentation.

Verify configuration:

docker run -v /dev/null:/var/run/nvidia-container-devices/all ubuntu:20.04 nvidia-smi -L

Expected output:

# GPU 0: NVIDIA L4 (UUID: GPU-3e71c48d-90c0-f46f-195b-4150320f9910)

---

Step 3: Install nvkind

# Install nvkind
go install github.com/NVIDIA/nvkind/cmd/nvkind@latest
cd nvkind
make

---

Step 4: Create nvkind Cluster

cat << EOF | nvkind cluster create --name=studio --config-template= -
kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
nodes:
- role: control-plane
- role: worker
  extraMounts:
    - hostPath: /dev/null
      containerPath: /var/run/nvidia-container-devices/all
EOF

---

Step 5: Set up the kubectl context

kubectl cluster-info --context kind-studio

---

Optional: If you have limited network bandwidth, you can pre-pull the container images. See image pre-puller details.

NAMESPACE=default ./deployment-scripts/images-pre-puller/deploy-image-prepuller.sh

Step 6: Install NVIDIA GPU Operator

helm repo add nvidia https://helm.ngc.nvidia.com/nvidia && \
helm repo update && \
helm install --wait --generate-name \
  -n gpu-operator --create-namespace \
  nvidia/gpu-operator --version=v25.10.0

GPU Operator

This may take several minutes. It manages NVIDIA drivers and device plugins in the cluster.

---

Step 7: Install Python Dependencies

pip install -r requirements.txt

---

Step 8: Deploy Geospatial Studio

./deploy_studio_k8s.sh

Deployment Time

~10-15 minutes depending on download speed and GPU operator initialization.

---

Monitor Deployment

Using k9s

k9s

---

Next Steps

Deployment Complete!

Your GPU-enabled Geospatial Studio is ready.

Continue to First Steps to access services and start using the studio.

---