[RFC] Allow to gracefully shutdown persistent tasks

[gajus]

Which project is this feature idea for?

Turborepo

Describe the feature you'd like to request

At the moment, as soon turbo receives SIGINT it kills the child process.

Describe the solution you'd like

I would like to be able to gracefully handle the shutdown.

Describe alternatives you've considered

Not using turbo

[gajus]

This issue is equivalent to pnpm/pnpm#2653 (comment)

[gajus]

Appears like this is a regression

#3711

[williamlmao]

Lately I have been having issues where hitting command + c will appear to stop the process in the terminal, but the app will still be running on the port. Is this the same issue you are referring to @gajus ?

[arlyon]

@williamlmao we had a regression in 1.8.4 released yesterday which was fixed in 1.8.5 released earlier today, could you try that?

[gajus]

I was using 1.8.3 when posting this, so probably a distinct issue from what @arlyon is referring to.

Mine is about waiting for the script to actually complete before force exiting turbo.

In ideal world, the way this would work is that the first time SIGINT is received, turbo passes the signal to the child processes and waits for them to exit, and only if another SIGINT is received that turbo force exits.

[chris-olszewski]

@gajus How are you invoking turbo? I've been digging into this and have been getting vastly different results depending on if you invoke turbo globally or via a package manager.

[gajus]

Yeah, pnpm has the same flaw

pnpm/pnpm#2653 (comment)

I was reporting what the behaviour is if directly invoking turbo though, i.e. ./node_modules/.bin/turbo

[BarakChamo]

Chiming in here with the same issue in a simple launch script that we need to run with turbo in persistent mode.
The reason for having a launch script is to run a docker-compose setup, but terminate it when turbo exits.

This is the result of running the script directly in the terminal from the monorepo with bun run up (at exit):

barakchamo@barak store % Stopping local Postgres
[+] Stopping 2/2
 ✔ Container store-pg_proxy-1  Stopped                                                                                                                                                                                                                                                                                             
 ✔ Container store-postgres-1  Stopped                                                                                                                                                                                                                                                                                             
canceled
[+] Running 3/0
 ✔ Container store-pg_proxy-1  Removed                                                                                                                                                                                                                                                                                             
 ✔ Container store-postgres-1  Removed                                                                                                                                                                                                                                                                                             
 ✔ Network store_default       Removed    

And this is the result of running the script from the monorepo root via turbo turbo run up (at exiting):

@this/store:up: Stopping local Postgres

barakchamo@barak bounce %  ERROR  run failed: command  exited (1)

There's no issue running the launch script with a package manager,
it's the execution from turbo that is causing the early termination before cleanup is completed.