docker-patterns

Define each service in docker-compose.yml with build targets, port mappings, and environment variables. Use bind mounts for hot reload, anonymous volumes to protect node_modules, and healthchecks with depends_on conditions so services start in the correct order.

Use separate stages for dependencies, development, build, and production. Pin base images to specific versions, create a non-root user, copy only production artifacts into the final stage, and add a HEALTHCHECK instruction for monitoring.

Services on the same Compose network resolve each other by service name as DNS hostnames. For isolation, define custom networks so, for example, a database is reachable only from the API tier and not from the frontend.

Docker Compose is designed for local development and single-host deployments. For production multi-container workloads, use an orchestrator such as Kubernetes, ECS, or Docker Swarm, which provide scaling, self-healing, and rolling updates.

Containers are ephemeral, so data written inside them disappears on removal. Mount a named volume such as pgdata to /var/lib/postgresql/data to persist database files across container restarts and rebuilds.

Load secrets from a gitignored .env file via env_file or inherit them from the host environment. Never hardcode credentials in the Dockerfile or compose file; in Swarm mode, use Docker secrets backed by files.