deploy-registry-server-with-otel

name	deploy-registry-server-with-otel
description	Deploy the ToolHive Registry Server to a Kind cluster with telemetry enabled. Use when you need to deploy the registry server for testing with OTEL metrics and tracing.
allowed-tools	Bash, Read
argument-hint	[local|latest]

Deploy Registry Server with Telemetry

Deploy the ToolHive Registry Server to a Kind cluster with OpenTelemetry telemetry enabled.

Arguments

• local - Build the image locally using ko and deploy (default)
• latest - Use the latest published image from GitHub

Prerequisites

Before running this skill, ensure:

• Kind cluster exists (run /deploy-otel first)
• kconfig.yaml file exists in the project root
• OTEL stack is deployed to the cluster

This skill deploys a minimal, single-user Postgres pod for the registry server's database — it has no persistence or HA guarantees. Use /deploy-registry-server-with-cnpg instead if you need a production-like database setup.

Steps

1. Verify Prerequisites

echo "Checking prerequisites..."

# Check kconfig.yaml exists
if [ ! -f kconfig.yaml ]; then
  echo "ERROR: kconfig.yaml not found. Run /deploy-otel first to create the Kind cluster."
  exit 1
fi

# Check kind cluster is accessible
if ! kubectl cluster-info --kubeconfig kconfig.yaml >/dev/null 2>&1; then
  echo "ERROR: Cannot connect to Kind cluster. Run /deploy-otel first."
  exit 1
fi

# Check OTEL collector is deployed (the collector's DaemonSet name carries a
# chart-generated "-agent" suffix, so match by label instead of a fixed name)
if ! kubectl get daemonset -n monitoring -l app.kubernetes.io/name=opentelemetry-collector --kubeconfig kconfig.yaml 2>/dev/null | grep -q otel-collector; then
  echo "WARNING: OTEL Collector not found in monitoring namespace."
  echo "Telemetry data will not be collected. Run /deploy-otel first."
fi

echo "Prerequisites verified."

2. Deploy Postgres

The registry server requires a Postgres database. This step deploys the minimal single-user Postgres pod defined in examples/otel/postgres.yaml into toolhive-system if one isn't already running.

echo "Checking for existing Postgres deployment..."
kubectl create namespace toolhive-system --kubeconfig kconfig.yaml --dry-run=client -o yaml | kubectl apply -f - --kubeconfig kconfig.yaml

if kubectl get deployment postgres -n toolhive-system --kubeconfig kconfig.yaml >/dev/null 2>&1; then
  echo "Postgres already deployed."
else
  echo "Deploying Postgres..."
  kubectl apply -f examples/otel/postgres.yaml --kubeconfig kconfig.yaml

  echo "Waiting for Postgres to be ready..."
  kubectl rollout status deployment/postgres -n toolhive-system --kubeconfig kconfig.yaml --timeout=90s
fi

3. Deploy Registry Server

Based on the argument, either build locally or use the latest image. Telemetry and the database connection are configured dynamically via Helm --set flags.

DEPLOY_MODE="${ARGUMENTS:-local}"

# Common telemetry configuration flags
TELEMETRY_FLAGS=(
  --set config.telemetry.enabled=true
  --set config.telemetry.serviceName=thv-registry-api
  --set config.telemetry.endpoint=otel-collector-opentelemetry-collector.monitoring.svc.cluster.local:4318
  --set config.telemetry.insecure=true
  --set config.telemetry.metrics.enabled=true
  --set config.telemetry.tracing.enabled=false
  --set config.telemetry.tracing.sampling=1.0
)

# Database configuration flags (points at the Postgres pod deployed in step 2)
DATABASE_FLAGS=(
  --set config.database.host=postgres.toolhive-system.svc.cluster.local
  --set config.database.user=thv_user
  --set config.database.database=toolhive_registry
  --set config.database.sslMode=disable
  --set-json 'extraEnv=[{"name":"THV_REGISTRY_DATABASE_PASSWORD","valueFrom":{"secretKeyRef":{"name":"db-credentials","key":"password"}}},{"name":"THV_REGISTRY_DATABASE_MIGRATIONPASSWORD","valueFrom":{"secretKeyRef":{"name":"db-credentials","key":"migration-password"}}}]'
)

if [ "$DEPLOY_MODE" = "local" ] || [ -z "$DEPLOY_MODE" ]; then
  echo "Building registry server image locally with ko..."

  # Check ko is installed
  if ! command -v ko >/dev/null 2>&1; then
    echo "ERROR: ko is not installed. Install from https://ko.build/"
    exit 1
  fi

  # Build the image
  REGISTRY_SERVER_IMAGE=$(KO_DOCKER_REPO=kind.local ko build --local -B ./cmd/thv-registry-api | tail -n 1)
  echo "Built image: $REGISTRY_SERVER_IMAGE"

  # Load into kind
  echo "Loading image into Kind cluster..."
  kind load docker-image "$REGISTRY_SERVER_IMAGE" --name thv-registry

  # Deploy with helm
  echo "Deploying registry server with telemetry enabled..."
  helm upgrade --install toolhive-registry-server deploy/charts/toolhive-registry-server \
    --namespace toolhive-system \
    --create-namespace \
    --kubeconfig kconfig.yaml \
    --set image.registryServerUrl="$REGISTRY_SERVER_IMAGE" \
    --set image.pullPolicy=Never \
    "${TELEMETRY_FLAGS[@]}" \
    "${DATABASE_FLAGS[@]}"

elif [ "$DEPLOY_MODE" = "latest" ]; then
  echo "Deploying latest registry server image from GitHub..."
  helm upgrade --install toolhive-registry-server deploy/charts/toolhive-registry-server \
    --namespace toolhive-system \
    --create-namespace \
    --kubeconfig kconfig.yaml \
    "${TELEMETRY_FLAGS[@]}" \
    "${DATABASE_FLAGS[@]}"
else
  echo "ERROR: Unknown deploy mode '$DEPLOY_MODE'. Use 'local' or 'latest'."
  exit 1
fi

4. Wait for Deployment

echo "Waiting for registry server to be ready..."
kubectl rollout status deployment/toolhive-registry-server -n toolhive-system --kubeconfig kconfig.yaml --timeout=2m

5. Verify Deployment

echo "Verifying deployment..."
kubectl get pods -n toolhive-system --kubeconfig kconfig.yaml

6. Show Telemetry Configuration

echo "Checking telemetry configuration..."
kubectl get configmap toolhive-registry-server-config -n toolhive-system -o jsonpath='{.data.config\.yaml}' --kubeconfig kconfig.yaml | grep -A 15 "telemetry:"

7. Provision the Grafana Dashboard

Grafana's sidecar auto-discovers dashboards from ConfigMaps labeled grafana_dashboard=1 in the monitoring namespace. Load the registry server dashboard this way:

echo "Provisioning the registry server Grafana dashboard..."
kubectl create configmap registry-server-dashboard \
  -n monitoring \
  --from-file=registry-server.json=examples/otel/grafana/dashboards/registry-server.json \
  --kubeconfig kconfig.yaml \
  --dry-run=client -o yaml | kubectl label -f - grafana_dashboard=1 --local -o yaml --kubeconfig kconfig.yaml | kubectl apply -f - --kubeconfig kconfig.yaml

echo "Dashboard provisioned. The Grafana sidecar picks it up within ~1 minute."

8. Display Access Instructions

cat <<'EOF'

=== Registry Server Deployment Complete ===

To access the registry server API:
  kubectl port-forward -n toolhive-system svc/toolhive-registry-server 8080:8080 --kubeconfig kconfig.yaml

Then test with:
  curl http://localhost:8080/registry/default/v0.1/servers

To check health and metrics (internal port, no auth):
  kubectl port-forward -n toolhive-system svc/toolhive-registry-server 8081:8081 --kubeconfig kconfig.yaml
  curl http://localhost:8081/health
  curl http://localhost:8081/metrics

View the dashboard in Grafana:
  kubectl port-forward -n monitoring svc/kube-prometheus-stack-grafana 3000:80 --kubeconfig kconfig.yaml
  Open http://localhost:3000/d/toolhive-registry/toolhive-registry-server (admin / admin)

EOF

Telemetry Configuration

The skill dynamically configures telemetry via Helm --set flags:

Setting	Value
config.telemetry.enabled	true
config.telemetry.serviceName	thv-registry-api
config.telemetry.endpoint	otel-collector-opentelemetry-collector.monitoring.svc.cluster.local:4318
config.telemetry.insecure	true
config.telemetry.metrics.enabled	true
config.telemetry.tracing.enabled	false
config.telemetry.tracing.sampling	1.0

Cleanup

To uninstall the registry server and its Postgres pod:

helm uninstall toolhive-registry-server --namespace toolhive-system --kubeconfig kconfig.yaml
kubectl delete -f examples/otel/postgres.yaml --kubeconfig kconfig.yaml --ignore-not-found
kubectl delete configmap registry-server-dashboard -n monitoring --kubeconfig kconfig.yaml --ignore-not-found