Node Ace instance does not self exit inside docker container

[jakebknowles-simpro]

When running agent workflows through something like claude and the node instance is inside a docker container, when I or Claude run node ace for any of the commands it does not self exit.

Because of this Claude gets hung up on the tests running and does not proceed as its awaiting an exit.

Couldn't find a solution to this online, other than manually exiting through my own code and have confirmed with other developers they have the same issue.

To recreate, just run a node instance in docker with adonis repo, run any ace command and you will see that it does not exit once finished.

If this is purposeful could someone point me towards the reason as to why, I can't really think of any.

Running the latest version of adonisjs.

This is the docker volume config

  backend:
    build:
      context: ./backend
      dockerfile: Dockerfile.dev
    container_name: my_backend
    ports:
      - "3333:3333"
      - "9229:9229" # Node.js debugger
    volumes:
      - ./backend:/app
      - backend_node_modules:/app/node_modules
      - ./backend/.env.docker:/app/.env:ro
    networks:
      - default
      - secretNetwork
    depends_on:
      postgres:
        condition: service_healthy

[jakebknowles-simpro]

Hey this is a serious problem with your framework!

[thetutlage]

Hey, we would require a proper reproduction repo + the Dockerfile in order to look into the issue

[jakebknowles-simpro]

Consider it done, I have a fix as well, I see you have a few requirements on PRs. Happy to follow those, I should be able to get the correct structure in place for this issue over the weekend.

My company is moving very fast on a new system that will be used at a very large scale using your framework, I know they don't pay you for this, but I'd be eternally grateful if you can keep an eye on this space.

Thank you!

[jakeb-k]

Made an issue with the requirements you listed.
adonisjs/ace#166

[thetutlage]

Thanks. Will look into it!

[MukundaKatta]

Adonis's HTTP server keeps the event loop alive until you explicitly httpServer.close() — that's by design for long-running servers, but ace test / ace db:seed scripts should process.exit(0) at the end. Usual culprit is an orphan DB connection (Lucid/knex pool) or a redis client that wasn't .close()'d. Add await db.manager.closeAll() (or equivalent for your provider) at the end of the ace handler, and the instance will exit cleanly — especially useful in Docker / CI where stdin isn't a TTY.

[jakebknowles-simpro]

Thank you! I apologise for being a noob