Self-host

Run the dashboard on your own infrastructure. The same build targets three runtimes — pick one. Every target reads the same CLICKHOUSE_* environment, so config carries over if you switch later.

The dashboard is stateless — it only reads ClickHouse system tables — so scaling out and restarting are always safe.

Cloudflare Workers

Edge hosting. Builds with the OpenNext adapter and deploys to a Worker. Best when you want a globally cached, serverless front end with no servers to run.

# 1. Clone and install
git clone https://github.com/duyet/clickhouse-monitoring.git
cd clickhouse-monitoring
bun install

# 2. Put your ClickHouse credentials in .env.prod (or .env.local)
cat <<'EOF' > .env.prod
CLOUDFLARE_API_TOKEN=<token-with-workers-deploy-permission>
CLICKHOUSE_HOST=https://clickhouse.example.com:8443
CLICKHOUSE_USER=default
CLICKHOUSE_PASSWORD=change-me
EOF

# 3. Push secrets to the Worker, then build + deploy + populate cache
bun run cf:config
bun run cf:deploy

cf:deploy runs cf:build → wrangler deploy --minify → populateCache remote. Without CLOUDFLARE_API_TOKEN it falls back to wrangler login. Full reference: Cloudflare.

Docker

A single self-hosted container. Use the published image from GHCR — pin a release tag in production. Fastest path if you already run ClickHouse somewhere.

docker run -d --name chmonitor -p 3000:3000 \
  -e CLICKHOUSE_HOST='https://clickhouse.example.com:8443' \
  -e CLICKHOUSE_USER='default' \
  -e CLICKHOUSE_PASSWORD='change-me' \
  ghcr.io/duyet/chmonitor:latest

# Open http://localhost:3000

To try it with a throwaway ClickHouse alongside the app, use the bundled Compose file instead:

git clone https://github.com/duyet/clickhouse-monitoring.git
cd clickhouse-monitoring
docker compose up -d        # starts ClickHouse + the dashboard

If ClickHouse runs on the Docker host, localhost inside the container points at the container — use host.docker.internal (add --add-host=host.docker.internal:host-gateway on Linux). Full reference: Docker.

Kubernetes + Helm

Cluster-managed deployment. The Helm chart is vendored in-repo at deploy/helm/chmonitor so it tracks the app it ships. It stores the password in a Secret, runs as the non-root app user, and wires liveness/readiness probes.

# 1. Get the chart
git clone https://github.com/duyet/clickhouse-monitoring.git
cd clickhouse-monitoring

# 2. Install, passing your ClickHouse connection
helm install my-chm ./deploy/helm/chmonitor \
  --set clickhouse.host="https://clickhouse.example.com:8443" \
  --set clickhouse.user="default" \
  --set clickhouse.password="change-me"

# 3. Reach it locally (service is <release>-chmonitor)
kubectl port-forward svc/my-chm-chmonitor 3000:3000
# Open http://localhost:3000

Upgrade or remove later:

helm upgrade my-chm ./deploy/helm/chmonitor -f my-values.yaml
helm uninstall my-chm

For production, keep the password out of --set (use clickhouse.existingSecret or a secrets operator), enable an Ingress, and turn on the HPA via autoscaling.enabled. Raw kustomize manifests and full options: Kubernetes.

• Cloudflare
• Docker
• Kubernetes
• Production Checklist