How do I chose between upstart, runit, supervisor, daemontools, ... for restarting a process if it dies? (process supervision / monitoring)

Question

I have a Nginx web proxy, gunicorn web server, and a python/flask web app. The Gunicorn process apparently died, and I want to ameliorate that in the future by looking into a utility that can monitor and restart the gunicorn process in the event it crashes again.

I've found several process supervision utilities that can do the job:

• daemontools
• launchd
• runit
• s6
• supervisor
• SystemD
• upstart
• ...

Is there a comprehensive article that compares and contrasts the various utilities used to monitor and restart a process?

https://en.wikipedia.org/wiki/Process_supervision

Answer 1

• runit is a successor to daemontools (both are written in c)

• supervisord uses python.

I've been using runit with socklog by the same author inside Alpine Linux lxc containers for around 10 months to manage web / database & various other services. It is light, easy to manage & I have had no service failures. The logging daemon also runs as it's own user & not root which is nice.

voidlinux uses runit as it's init system & also for service supervision (search the package tree for run files for examples of runit scripts).

Answer 2

If your distro uses Upstart, go with it. It has very basic support for job restarting, but includes limits that can prevent from restart loop, as mentioned by @EEAA.

If your OS uses another init program, don't change it. I can't really help you with the other tools you mentioned, as I generally use Ubuntu where Upstart is still present (as of the latest LTS), so I have little to do with them. But it's not a hard task to create a simple script which is run from cron once a minute (or more frequently in ie. a loop), which can check if a PID exists and issue restart on failure.