When using docker-compose run for one-off tasks (like database migrations or batch jobs), leftover containers accumulate unless manually removed. While the CLI's --rm flag solves this, embedding this behavior directly in docker-compose.yml maintains consistency across teams and CI/CD pipelines.

As of Docker Compose v2.3+ (Compose Specification), you can define this at the service level:

services:
  migrations:
    image: postgres:alpine
    profiles: ["tools"]
    labels:
      com.docker.compose.autoremove: "true"  # Key configuration
    command: ["sh", "-c", "psql -U $$DB_USER -d $$DB_NAME -f /migrations/*.sql"]

For older versions or more control, combine healthchecks with depends_on:

services:
  batch-job:
    image: python:3.9
    healthcheck:
      test: ["CMD", "test", "-f", "/tmp/.finished"]
      interval: 5s
      retries: 1
    labels:
      on-success: "remove"
    command: ["python", "process_data.py"]

For a Node.js test runner that should self-destruct:

version: '3.8'
services:
  tester:
    build: .
    autoremove: true  # Works in some community plugins
    environment:
      - NODE_ENV=test
    volumes:
      - ./tests:/app/tests
    command: ["npm", "test"]

• The com.docker.compose.autoremove label requires Docker Engine 20.10+
• For production containers, consider restart: on-failure instead
• Always test in your specific environment - some Docker versions implement this differently

When using docker-compose run for temporary service operations, manually adding the --rm flag every time becomes tedious. Developers need a way to enforce this behavior directly in the Compose file.

As of Docker Compose v2.x, there's no native auto_remove or rm option that can be specified under a service in docker-compose.yml. This differs from the docker run --rm functionality available in plain Docker.

# What we WANT (but doesn't work):
services:
  my_service:
    image: alpine
    rm: true  # Invalid Compose syntax

Option 1: Using Compose Overrides

Create a separate override file that modifies the runtime configuration:

# docker-compose.override.yml
services:
  my_service:
    command: sh -c "your_command && exit"
    restart: "no"

This won't automatically remove containers but prevents them from restarting.

Option 2: Custom Entrypoint Script

Create an entrypoint that ensures container termination:

# Dockerfile
FROM alpine
COPY entrypoint.sh /entrypoint.sh
RUN chmod +x /entrypoint.sh
ENTRYPOINT ["/entrypoint.sh"]

#!/bin/sh
# entrypoint.sh
trap "exit 0" SIGTERM
"$@"

Option 3: Using Healthchecks

Configure a healthcheck that forces container exit:

services:
  my_service:
    image: alpine
    healthcheck:
      test: ["CMD", "exit", "0"]
      interval: 1s
      retries: 1
    restart: "no"

For development environments, consider using project-level defaults in ~/.docker/config.json:

{
  "compose": {
    "run": {
      "auto-remove": true
    }
  }
}

Combine this with shell aliases for maximum efficiency:

alias dcrun='docker compose run --rm'

The Docker community has discussed adding native support for this feature. Track GitHub issue #6476 for updates. Meanwhile, the workarounds above provide functional alternatives.