Hacker News new | past | comments | ask | show | jobs | submit
Yep, that's exactly the mess I'm talking about. Scripts that are ALWAYS incorrect, incomplete, and unreliable.

For example, can it kill a runaway Apache process tree? Or safely confine the CGI scripts so that they don't cause the OOM killer to terminate the sshd?

And what happens if the "stop" command hangs indefinitely?

> And no need to track pid. It uses pgrep with the flags and the program name by default. I believe OpenRC does the same.

Facepalm.

> Scripts that are ALWAYS incorrect, incomplete, and unreliable.

The BSDs have used it for years now and it works well enough

> Or safely confine the CGI scripts so that they don't cause the OOM killer to terminate the sshd?

That’s a different thing than service management. It’s handled by the login class of the program’s user which specify such things.

> And what happens if the "stop" command hangs indefinitely?

The logic for stop is in the rc.subr file. There’s a timeout for trying to send TERM, then it’s KILL.

> Facepalm.

How many services do you intend to run on a single machine/vm? rc is simple, use existing tooling, and it’s more of convenience than a new things you have to take care of. I run debian and while systemd works, I’m not sure I need all of those stuff.

loading story #50476390