> Dnia wtorek, 22 maja 2012 00:36:33 Łukasz Mierzwa pisze: >> Dnia poniedziałek, 21 maja 2012 22:14:19 Roberto De Ioris pisze: >> > I have checked the code, it looks like waitpid() calls are blocked >> while >> > in cheap (no workers) mode. Could it be your situation ? >> >> Do I read it correctly? cheap == using --idle option and having all >> workers >> shut down due to no requests? If so than yes, this is testing cluster >> and >> the app in question did not had much requests over last week, so most of >> the time it was running with no active workers. Should I disable --idle >> option and retest? >> >> > Reapers are called by workers, while command-crons (the ones you >> define >> > via --cron) are generated by the master. >> >> It wouldn't help my testing cluster, but my production apps are nagios >> checked, so there is at least one request per minute and --idle can't >> kill >> all workers. > > After night without --idle option I don't have any zombie processes. It > doesn't matter if I also use --reaper, removing --idle alone is enough (at > least for ~8 hours). > >
Can you retry with the latest tip ? Now, during cheap mode, the master should continue calling waitpid() periodically -- Roberto De Ioris http://unbit.it _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
