Skip to main content

Multi-Service Apps

Kindling handles multi-service architectures natively. Each service gets its own DevStagingEnvironment — independent deployment, scaling, health checks, and dependencies — while sharing the same cluster, registry, and ingress controller.

Example: microservices demo​

The built-in microservices example has four services:

microservices/
├── gateway/     → API gateway, port 9090, ingress on gateway.localhost
├── orders/      → Order service, port 8080, depends on postgres + redis
├── inventory/   → Inventory service, port 8080, depends on mongodb
└── ui/          → React frontend, port 3000, ingress on ui.localhost

Try it​

# Copy the demo
cp -r ~/.kindling/examples/microservices ~/kindling-demo
cd ~/kindling-demo

# Create a GitHub repo
gh repo create kindling-demo --public --source . --push

# Connect runners
kindling runners -u <your-user> -r <your-user>/kindling-demo -t <pat>

# Push to trigger the workflow
git push

How multi-service workflows work​

Each service gets a build step and a deploy step in the workflow:

- name: Build gateway
  uses: kindling-sh/kindling/.github/actions/kindling-build@main
  with:
    name: gateway
    context: ${{ github.workspace }}/gateway
    image: "registry:5000/gateway:${{ env.TAG }}"

- name: Deploy gateway
  uses: kindling-sh/kindling/.github/actions/kindling-deploy@main
  with:
    name: "${{ github.actor }}-gateway"
    image: "registry:5000/gateway:${{ env.TAG }}"
    port: "9090"
    ingress-host: "${{ github.actor }}-gateway.localhost"

Services with dependencies declare them in the deploy step:

- name: Deploy orders
  uses: kindling-sh/kindling/.github/actions/kindling-deploy@main
  with:
    name: "${{ github.actor }}-orders"
    image: "registry:5000/orders:${{ env.TAG }}"
    port: "8080"
    dependencies: |
      - type: postgres
        version: "16"
      - type: redis

The operator auto-provisions each dependency, creates a Service for it, and injects connection strings (DATABASE_URL, REDIS_URL) into the app container.

Resource considerations​

Each service adds a pod + its dependencies. On a Mac laptop:

SetupTypical memory
2–3 lightweight services, no deps~1.5 GB total
4 services + postgres + redis + mongodb~2–3 GB total
6+ services with heavy deps (Kafka, Elasticsearch)4–8 GB total

See Docker Desktop Resources for allocation recommendations.

Live sync across services​

You can sync individual services independently:

# Sync the gateway service
kindling sync -d gateway --src ./gateway --restart

# In another terminal, sync the orders service
kindling sync -d orders --src ./orders --restart

Shared dependencies​

If two services both need Postgres, each gets its own independent instance. This mirrors production isolation and avoids schema conflicts. The operator names them <dse-name>-postgres so they don't collide.

Using kindling generate for multi-service repos​

kindling generate -k sk-... -r /path/to/multi-service-repo

The scanner detects each directory with a Dockerfile, analyzes docker-compose.yml for depends_on relationships, and generates a workflow with the correct build order and dependency declarations.