Hacker News Reader

Comment by amq | Hacker News Reader

Would it be a suitable replacement for supervisord in containers?

Most likely. It can even run as an unprivileged user (and users can manage their own daemons with it).

Though in most cases I find that people just want the "zombie problem" to go away. I wrote a simple init[1] that implements all of the key pieces of an init and signal forwarder.

[1]: https://github.com/cyphar/initrs

I see you mentioned tini (another such init) in the README as being "not as simple as it should be". We're using https://github.com/Yelp/dumb-init at work; if you know it, what do you think of it?

dumb-init is better, but they don't use socketfd which means they might drop signals in certain cases (in the race between consecutive sigwaits). I will concede this is a minor point, and I still have to finish implementing all of the necessary signal semantics in initrs, but it is kinda important.

Also I'm learning Rust and it was a good opportunity to practice by implementing something I already knew how to easily do in C.

Any reason you didn't go with something like dumb-init[1]?

[1] https://github.com/Yelp/dumb-init

See https://news.ycombinator.com/item?id=14478725

I'm waiting for the same. Supervisord seems to be dead project now, without any releases since the last 10 months or so.

I'm not sure what their release process looks like, but the Github for supervisor seems reasonably active: https://github.com/Supervisor/supervisor

Sadly it doesn't seem to play well when run under Python 3. Not a big deal now, but hopefully one of these years, desktop Linux distributions will have a 3.x release of Python as the default.

Reply on news.ycombinator.com