Dnia środa, 30 listopada 2011 19:45:52 Łukasz Mierzwa pisze: > Dnia środa, 30 listopada 2011 19:33:49 Roberto De Ioris pisze: > > They are related, you have to add > > > > memory-report = true > > > > to have those flags do their job. > > > > Yes, it is a bit annoying, i will try to found a way to fix it without > > adding the memory reports in the logs. > > I forgot to add my default settings, memory-report is enabled and it is > normally working (it was working before and it works on the other server): > > [uwsgi] > no-default-app = true > master = true > memory-report = true > harakiri = 120 > harakiri-verbose = true > idle = 300 > cheap = true > vacuum = true > no-orphans = true > reload-on-rss = 256 > max-requests = 1000 > reload-mercy = 30 > post-buffering = 8192 > emperor = /etc/uwsgi/emperor > > I'm running emperor using "uwsgi --inherit /etc/uwsgi/defaults.ini" command > line.
I've noticed that after worker is reloaded due to max-requests, it eats as much memory as before reload. Reloads caused by editing vassal config file makes all workers do full restart and they eat very little memory. Should max requests/max memory limit reloads be technically the same as config edit reloads? I think that this is why I got this "worker reload bomb", my workers reached max memory limit but reload did not caused them to release memory, so they were reloaded at every request. I just disabled KSM on one node to see if it makes any difference. Łukasz Mierzwa _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
