Comments on: Upstart 0.5: Job Lifetime

By: Juri Haberland

Juri Haberland — Fri, 31 Oct 2008 16:26:45 +0000

Is there a possibility to configure upstart to retry a start of a failed daemon after a couple of minutes, like INIT does, if a service is respawning to fast? Currenty upstart just disables the service forever, if it is respawning to fast.

By: esp

esp — Sat, 12 Jul 2008 19:47:23 +0000

the “respawn” functiopn i didnt know of, Is it somewhere in the doc?

I cant get it to work, I have a daemon here wich keeps crashing, i dont know why.
I made a upstart script for it, so that i could restart it with the respawn, but that doesnt work.
If i check with ’status jobname’ after some hours it says ‘jobname (stop) waiting’

I would appreciate input on how to get that daemon running forever.

By: chrisfarms

chrisfarms — Wed, 09 Jul 2008 10:32:29 +0000

I’m really loving upstart, great work. wish more of my boxes had 0.5 on them though.
Can’t wait until all the rc structures and init.d are gone.

I think the only thing that makes it tricky to write scripts is the lack of a list of events and what emits them. I understand that this would require some kind of registration system, which does sway a tad from upstart’s simplicity…. maybe just a log of every unique event ever emitted would be handy.

By: Scott James Remnant

Scott James Remnant — Sun, 20 Apr 2008 20:25:59 +0000

@kees: good point

@jeff: probably a d-bus command

@sitsofe: depends whether it has inetd support

@diego: why? if any package has the right to claim generic names like start, stop and status – it’s PID #1 :p

By: Diego

Diego — Sun, 20 Apr 2008 17:28:58 +0000

“status” is not a very good name for a executable, I must say :/

By: Sitsofe

Sitsofe — Sun, 20 Apr 2008 10:07:24 +0000

very of offtopic but will upstart ever do port reservation for a service it will later start like like launchd does (thus preventing issues like https://bugzilla.redhat.com/show_bug.cgi?id=430825 )?

By: Jeff Bailey

Jeff Bailey — Sat, 19 Apr 2008 23:47:18 +0000

Will up HUP or some such cause the respawns-exceeded timeout to get overridden? That would be nice for troubleshooting.

By: Kees Cook

Kees Cook — Sat, 19 Apr 2008 22:43:57 +0000

AIUI, Debian policy for maintainer scripts calling “start” on an init script requires that “start” exit 0 when the service is already started. Will there be something similar to start-stop-daemon’s “–oknodo” option?

By: Scott James Remnant

Scott James Remnant — Sat, 19 Apr 2008 20:46:42 +0000

@pkern: I was kinda picking on sobby as an example

By: Philipp Kern

Philipp Kern — Sat, 19 Apr 2008 19:26:39 +0000

We would be very curious to see backtraces of crashes of Sobby 0.4.x…