Multi-Container Apps | Compose to Production Patterns

You will assemble a production-shaped app: a reverse proxy in front of an API, a background worker, and a database with persistent storage. By the end you will run a realistic topology with config via environment and secrets.

Learning Objectives

Compose a 4-tier app (proxy, API, worker, DB).
Inject configuration with environment and secrets.
Scale a stateless service horizontally.
Time: ~4 hours · Difficulty: Intermediate · Prereqs: the Compose lab completed.

Architecture Overview

Environment Setup

You will need: Docker Compose v2 and the proxy/API/worker images (or simple placeholders).

Before you begin: never bake secrets into images; pass them at runtime.

Step-by-Step Execution

Define the multi-service stack

Each tier is a service; the proxy is the only one exposed to the host.

docker compose -f compose.yaml config

Validates and renders the merged configuration.

Provide secrets at runtime

printf 'supersecret' | docker secret create db_password - 2>/dev/null || echo "use env_file for compose"

Bring it up and scale the API

$ docker compose up -d --scale api=3

[+] Running 6/6 proj-api-1 Started proj-api-2 Started proj-api-3 Started

Progress So Far

graph LR A[01 Define stack] -->|done| B[02 Secrets] B -->|done| C[03 Up + scale] style A fill:#1a4a1a,stroke:#00ff00,color:#fff style B fill:#1a4a1a,stroke:#00ff00,color:#fff style C fill:#1a4a1a,stroke:#00ff00,color:#fff

Testing & Validation

for i in 1 2 3; do curl -s http://localhost/api/whoami; done

You should see responses from different API replicas, confirming load distribution behind the proxy.

Troubleshooting

Only one replica answers: ensure the proxy load-balances across the service name, not a fixed container.
Worker idle: confirm it consumes the same Redis queue the API publishes to.
Secret not found: Compose uses secrets: with files; Swarm uses docker secret.

Extension Ideas

Translate this topology to Kubernetes Deployments and Services.
Add autoscaling and resource limits.
Front it with the Cloudflare Tunnel pattern.

Key Results

Ran a 4-tier app with only the proxy exposed.
Injected configuration and secrets at runtime, not in images.
Scaled the stateless API to 3 replicas behind the proxy.
Verified load distribution across replicas.

Build a Realistic Multi-Container Application