On Thu, Feb 21, 2013 at 1:15 PM, Chris McDonough <chr...@plope.com> wrote:
> On Thu, 2013-02-21 at 13:07 -0800, David Birdsong wrote: > > > > > > On Thu, Feb 21, 2013 at 12:57 PM, John Wong <gokoproj...@gmail.com> > > wrote: > > Hi Dan, > > > > > > The fix didn't work. It still leave one child behind. > > Like I said, this is my first time encountering the bug and we > > didn't change any code at all. > > Strangely, I have another guniconr_paster app running and it > > can kill without problem. > > > > > > > > > > supevisord only kills it's direct children, it doesn't go after any > > grandchildren. > > > > > You can tell it to clean up processes left behind by ill-behaved > children by using the "stopasgroup" option in the process config, I > think. It was added pretty recently (in supervisor 3.0b1) > oh, cool. thanks for the correction. i guess signaling a process group makes sense. i'm stuck on a few revisions behind since my patch to include various RLIMIT configurations is kinda crappy but we rely on it very heavily. > - C > > > > > Thanks. > > > > > > John > > > > On Thu, Feb 21, 2013 at 1:36 PM, Dan Crosta <d...@magnetic.com> > > wrote: > > We use supervisor with gunicorn with stopsignal=INT > > successfully -- might be worth a try. > > > > - Dan > > > > On Feb 21, 2013, at 1:09 PM, "David Birdsong" > > <david.birds...@gmail.com> wrote: > > > > > if gunicorn is doing it's own forking, which i'm > > guessing it does, > > > then supervisord won't know about the child > > processes. > > > > > > i've not used gunicorn, but when i run tornado > > servers, i dont use > > > tornado's process manager, i instead use > > supervisord's argument > > > templating and numprocs to fork direct children from > > supervisord. the > > > arg templating empowers you to pass in different > > ports to each > > > process. > > > > > > so instead you'd run two gunicorn processes directly > > from supervisord > > > on different ports. > > > > > > > > > (aside, are you using gunicorn to run graphite > > instead of django?) > > > > > > On Thu, Feb 21, 2013 at 9:54 AM, John Wong > > <gokoproj...@gmail.com> wrote: > > >> Hi, > > >> > > >> I bumped into this biazzard problem today. I never > > had any issue with > > >> supervisord for the last two weeks since I started > > using it. > > >> > > >> > > > > (gcs)giabadmin@giab-master:/opt/graphyte/gcs/tests/integration$ > ps -elf|grep > > >> "gcs" > > >> 4 S 1011 3654 2899 1 80 0 - 3811 poll_s > > 17:48 ? 00:00:01 > > >> /opt/graphyte/vens/gcs/bin/python > /opt/graphyte/vens/gcs/bin/gunicorn_paster > > >> /opt/graphyte/gcs/development.ini -w 1 -t 3600 > > >> 1 S 1011 3659 3654 8 80 0 - 11852 ep_pol > > 17:48 ? 00:00:11 > > >> /opt/graphyte/vens/gcs/bin/python > /opt/graphyte/vens/gcs/bin/gunicorn_paster > > >> /opt/graphyte/gcs/development.ini -w 1 -t 3600 > > >> 1 S postgres 3674 1347 0 80 0 - 13192 sk_wai > > 17:48 ? 00:00:00 > > >> postgres: postgres gcs 127.0.0.1(47334) idle > > >> 1 S postgres 3732 1347 0 80 0 - 13001 sk_wai > > 17:49 ? 00:00:00 > > >> postgres: postgres gcs 127.0.0.1(47419) idle > > >> 0 R 1011 3747 1105 3 80 0 - 1157 - > > 17:50 pts/0 00:00:00 > > >> grep --color=auto gcs > > >> > > >> > > >> We are looking at GCS. We have two of these. > > >> > > >> Then I ran super which is a shortcut for sudo > > supervisrdctl (I made this > > >> into a bash alias) > > >> > > >> > > > > (gcs)giabadmin@giab-master:/opt/graphyte/gcs/tests/integration$ > super stop > > >> gcs > > >> gcs: stopped > > >> > > >> I waited maybe a minute and I saw > > >> > > >> > > > > (gcs)giabadmin@giab-master:/opt/graphyte/gcs/tests/integration$ > ps -elf|grep > > >> "gcs" > > >> 1 S 1011 3659 1 4 80 0 - 11852 ep_pol > > 17:48 ? 00:00:11 > > >> /opt/graphyte/vens/gcs/bin/python > /opt/graphyte/vens/gcs/bin/gunicorn_paster > > >> /opt/graphyte/gcs/development.ini -w 1 -t 3600 > > >> 1 S postgres 3674 1347 0 80 0 - 13192 sk_wai > > 17:48 ? 00:00:00 > > >> postgres: postgres gcs 127.0.0.1(47334) idle > > >> 1 S postgres 3732 1347 0 80 0 - 13001 sk_wai > > 17:49 ? 00:00:00 > > >> postgres: postgres gcs 127.0.0.1(47419) idle > > >> 0 R 1011 3784 1105 0 80 0 - 1157 - > > 17:51 pts/0 00:00:00 > > >> grep --color=auto gcs > > >> > > >> Odd. Still alive. 3659 is still alive. > > >> > > >> I can't start my app until I manually kills 3659. > > Do we know what's causing > > >> the trouble? > > >> > > >> This is my conf: > > >> > > >> [program:gcs] > > >> command=/opt/graphyte/vens/gcs/bin/gunicorn_paster > > >> /opt/graphyte/gcs/development.ini -w 1 -t 3600 > > >> user=giabadmin > > >> autostart=true > > >> autorestart=false > > >> stopsignal=QUIT > > >> log_stdout=true > > >> log_stderr=true > > >> logfile=/var/log/graphyte/gcs2/supervisord.log > > >> stdout_logfile=/var/log/graphyte/gcs2/stdout.log > > >> stderr_logfile=/var/log/graphyte/gcs2/stderr.log > > >> logfile_maxbytes=20MB > > >> logfile_backups=10 > > >> > > >> Thanks and sorry for the long post. > > >> > > >> Cheers, > > >> John > > >> > > >> _______________________________________________ > > >> Supervisor-users mailing list > > >> Supervisor-users@lists.supervisord.org > > >> > > > https://lists.supervisord.org/mailman/listinfo/supervisor-users > > > _______________________________________________ > > > Supervisor-users mailing list > > > Supervisor-users@lists.supervisord.org > > > > > > https://lists.supervisord.org/mailman/listinfo/supervisor-users > > _______________________________________________ > > Supervisor-users mailing list > > Supervisor-users@lists.supervisord.org > > > https://lists.supervisord.org/mailman/listinfo/supervisor-users > > > > > > > > > > _______________________________________________ > > Supervisor-users mailing list > > Supervisor-users@lists.supervisord.org > > https://lists.supervisord.org/mailman/listinfo/supervisor-users > > > > > > _______________________________________________ > > Supervisor-users mailing list > > Supervisor-users@lists.supervisord.org > > https://lists.supervisord.org/mailman/listinfo/supervisor-users > > >
_______________________________________________ Supervisor-users mailing list Supervisor-users@lists.supervisord.org https://lists.supervisord.org/mailman/listinfo/supervisor-users
