Cheatsheet: Docker Compose

docker compose stop stops the running containers but leaves them, their networks, and their volumes in place, so a later docker compose start resumes quickly. docker compose down stops the containers and removes the containers and default network, but keeps named volumes (your data survives). docker compose down -v additionally removes named volumes declared under the volumes: key, so use -v only when you actually want to wipe the app's persisted data.

docker compose logs <service> shows that service's history; add -f (--follow) to keep streaming new lines, and -n 100 (--tail 100) to start from only the last 100 lines instead of the whole history. Combine them as docker compose logs -f --tail 50 <service>. Omit the service name to stream every service's logs interleaved, each prefixed with its service name.

exec runs a command inside an already-running container for that service (docker compose exec web bash), so it fails if the service isn't up yet. run starts a brand-new, one-off container from the service's image (docker compose run --rm web python manage.py migrate), which is what you want for setup scripts or commands that don't need the service's main container already running. run creates a new container each time (add --rm to auto-remove it afterward); it does not attach to a container up already started.

Compose automatically merges a docker-compose.override.yml file sitting next to your compose.yml if you run docker compose with no -f flags -- put dev-only overrides (a different port mapping, bind-mounted source code, debug env vars) there and keep it out of version control if it's personal. For explicit, named variants (staging.yml, prod.yml), pass multiple files in order: docker compose -f docker-compose.yml -f docker-compose.prod.yml up -d -- later files override matching keys from earlier ones.

Tag a service with profiles: ["debug"] (or similar) in the compose file, and Compose skips it by default -- normal docker compose up won't start it. Run docker compose --profile debug up to include every service tagged with that profile alongside the untagged (always-on) ones. This is the standard way to keep optional tooling (a debugger, an admin UI, seed-data jobs) out of the default startup without deleting it from the file.