CAST AI Castware Operator

An operator that manages Castware components.

How to run in a local environment

1. Generate an api key in console
2. Set the environment variable API_KEY with the generated key
3. Set the API_URL environment variable to the api url of the environment where you generated the api key
4. Run make run-local

Sample custom resources for cluster and component are generated in local/samples.

Local environment cleanup

1. Run make updeploy

Ignore the error about the missing castware-operator-controller-manager deployment.

How to run castware-operator in local kind cluster using local chart ./charts/castai-castware-operator

Initial Setup

1. Create Kind cluster:

   kind create cluster --name castware-operator

2. Set required environment variables:

   export API_KEY="your-api-key-from-console"
   export API_URL="https://api.cast.ai"  # or your environment URL
   export API_KEY_BASE64=$(echo -n "$API_KEY" | base64)

3. Install operator in Kind cluster:

   ./local/install-local.sh

   This script will:

   • Build the operator binary (make build)
   • Build Docker image (make docker-build)
   • Load image into Kind cluster
   • Install Helm chart with local image
   • Create default Cluster and Component resources

Development Workflow

After making code changes, quickly reload the operator without full reinstall:

./local/reload-operator.sh

This script will:

• Rebuild operator binary and Docker image
• Load new image into Kind cluster
• Restart the operator deployment
• Wait for rollout to complete

Benefits:

• ~30 seconds vs ~2 minutes for full reinstall
• Preserves existing CRs (Cluster and Components)
• No need to delete/recreate resources

Monitoring

# Watch pods
kubectl get pods -n castai-agent -w

# View operator logs
kubectl logs -n castai-agent -l app.kubernetes.io/name=castware-operator -f

# Check Cluster CR status
kubectl get cluster -n castai-agent -o yaml

# Check Helm release status
helm status castware-operator -n castai-agent

Cleanup

# Uninstall Helm release
helm uninstall castware-operator -n castai-agent

# Delete Kind cluster
kind delete cluster --name castware-operator

How to use local helm registry

1. Deploy ChartMuseum in the cluster

kubectl apply -f local/registry.yaml
kubectl wait --for=condition=ready pod -l app=chartmuseum -n registry --timeout=60s

2. Port-forward to access ChartMuseum from your machine

kubectl port-forward -n registry svc/chartmuseum 5001:8080 &

3. Add the local repo to Helm as castai-helm

helm repo add castai-helm http://localhost:5001/helm-charts
helm repo update

4. Package and push local charts

cd charts

# Package the chart (update version in Chart.yaml first if needed)
helm package ./castai-castware-operator

# Push to ChartMuseum
curl --data-binary "@castware-operator-<version_number>.tgz" http://localhost:5001/helm-charts/api/charts

# Verify
curl -s http://localhost:5001/helm-charts/api/charts | jq 'keys'

5. Mirror charts from public CAST AI repo (optional)

# Add public repo
helm repo add castai-public https://castai.github.io/helm-charts
helm repo update

# Pull and push to local registry
helm pull castai-public/castai-agent --version 0.127.0
helm pull castai/castai-spot-handler --version 0.29.0
curl --data-binary "@castai-agent-0.127.0.tgz" http://localhost:5001/helm-charts/api/charts
curl --data-binary "@castai-spot-handler-0.29.0.tgz" http://localhost:5001/helm-charts/api/charts		

6. Test self-upgrade

1. Push at least two versions of castware-operator to the local registry
2. Install the operator with the older version:

   helm upgrade --install castware-operator castai-helm/castware-operator \
       --namespace castai-agent --create-namespace \
       --set defaultCluster.helmRepoURL="http://chartmuseum.registry.svc.cluster.local:8080/helm-charts" \
       --version "0.1.0"
       (...)

3. The operator will use the in-cluster URL (chartmuseum.registry.svc.cluster.local:8080) for self-upgrade
4. Trigger upgrade via CAST AI action or manually update the Cluster CR