ok, i submitted a ulimit patch:
https://github.com/Supervisor/supervisor/pull/229


On Wed, Aug 22, 2012 at 3:52 PM, David Birdsong <[email protected]>wrote:

> i needed ulimit support pretty badly > 1 year ago, so i copy pasted
> code for the parent process into a program setup.
>
> was useful for all setrlimits:
>  - numprocs
>  - numfiles
>  - stack size
>
> the patch is really ugly and is hard to maintain against newer
> versions of supervisord, so i'd love for a better patch to surface.
>
> On Wed, Aug 22, 2012 at 10:20 AM, Jason Koppe <[email protected]> wrote:
> > I'm still keen on seeing ulimit support added.  Anyone else working on
> this
> > in a fork?
> >
> > On Thu, Sep 9, 2010 at 12:17 PM, Jason Koppe <[email protected]> wrote:
> >>
> >> We're using a wrapper around supervisorctl right now to accomplish
> custom
> >> ulimits, but this clearly isn't ideal.
> >>
> >>
> >> On Thu, Sep 9, 2010 at 11:40, Jason Koppe <[email protected]> wrote:
> >>>
> >>> I'd love ulimit or pam support, too!
> >>>
> >>> On Thu, Sep 9, 2010 at 11:34, Jordan Sissel <[email protected]> wrote:
> >>>>
> >>>> Howdy howdy,
> >>>>
> >>>> I've been using supervisor for a few weeks (moving from previously
> using
> >>>> daemontools), and I have a growing list of stuff I'd like to see in
> the
> >>>> project and would be happy to code and contribute myself - the docs
> say to
> >>>> email the list if I'm interested in contributing, so here's my
> feature list
> >>>> in no particular order:
> >>>>
> >>>> - ulimit support. supervisord doesn't invoke pam so it ignores
> >>>> /etc/security/limits.conf on Linux systems. Further, I would like to
> specify
> >>>> ulimit values per-program.
> >>>> - 'startretries=unlimited' would be excellent. I work around this by
> >>>> setting startretries=1000000, but it's not ideal since '1000000'
> doesn't
> >>>> explain my intent.
> >>>> - Want a built-in way to send signals. That is, I want 'supervisorctl
> >>>> signal <process> HUP' because many things support such signals for
> reloading
> >>>> config files, etc. Current workaround is to use 'supervisorctl pid
> <thing> |
> >>>> awk | xargs kill -SIGNAL' which isn't very awesome.
> >>>> - As far as I can tell, there is no way to force a process out of
> >>>> 'backoff' state. I have tried 'restart' and other commands. Having
> this
> >>>> ability would be good.
> >>>> - I also don't see a way to tune backoff timeouts, etc.
> >>>> - I want logs with timestamps. Not all programs output messages with
> >>>> timestamps. daemontools had 'multilog' for this kind of thing that
> would
> >>>> prefix process output with timestamps (among other things)
> >>>> - the event notification stuff could be nicer with a sample python
> >>>> library implementation to save some folks from the details of the wire
> >>>> protocol between supervisor and event handlers.
> >>>>
> >>>> Apologies if some of the above features are already implemented and
> >>>> documented. I tried to be thorough in my reading of the (very) awesome
> >>>> documentation this project has before making this list. :)
> >>>>
> >>>>
> >>>> -Jordan
> >>>>
> >>>> _______________________________________________
> >>>> Supervisor-users mailing list
> >>>> [email protected]
> >>>> http://lists.supervisord.org/mailman/listinfo/supervisor-users
> >>>>
> >>>
> >>>
> >>>
> >>> --
> >>> Jason Koppe
> >>> [email protected]
> >>> (210) 445-8242
> >>
> >>
> >>
> >>
> >> --
> >> Jason Koppe
> >> [email protected]
> >> (210) 445-8242
> >
> >
> >
> >
> > --
> > Jason Koppe
> > [email protected]
> > (210) 445-8242
> >
> > _______________________________________________
> > Supervisor-users mailing list
> > [email protected]
> > http://lists.supervisord.org/mailman/listinfo/supervisor-users
> >
>
_______________________________________________
Supervisor-users mailing list
[email protected]
https://lists.supervisord.org/mailman/listinfo/supervisor-users

Reply via email to