That does not help, unfortunately.
700 projects started in 6 minutes, tried even setting emperor-throttle to 1.
The CPU is almost idle during this process.
And unfortunately, it looks like none of the vassals work until emperor is
done spawning everyone. Not sure about this, but it looks like so.
Here's the config:
exec uwsgi --logto /var/log/uwsgi/emperor.log \
--die-on-term --emperor "/var/lib/crm/uwsgi_includes" \
--chmod-socket=777 --emperor-throttle=100 \
--emperor-on-demand-directory "/var/lib/crm/uwsgi_includes"
Thanks a lot!
On Thu, Oct 30, 2014 at 10:40 PM, Roberto De Ioris <[email protected]> wrote:
>
> > Hi, I have a pretty large emperor on-demand uwsgi setup, with 700
> projects
> > on one machine.
> >
> > When restarting the emperor, it takes somewhat long time for all the
> > sockets to start listening.
> > I'd say it takes around 0.5 seconds for every line like this to appear in
> > the logs, so the least lucky project has to wait several minutes to be
> > able
> > to get requests.
> >
> > [uwsgi-emperor] ....ini -> "on demand" instance detected, waiting for
> > connections on socket "...socket" ...
> >
> > Is there a way to speed this up somehow?
> >
> > Thanks!
> > _______________________________________________
> >
>
>
> One of the job of the Emperor is to avoid fork() bombing. It solves it
> using a throttling system that slow down spawning upto 1000 milliseconds
> (1 second).
>
> You can tune that value with --emperor-throttle <n>
>
> (<n> is in milliseconds)
>
> Try experimenting with lower values (i would start with 100)
>
> Let me know
>
>
> --
> Roberto De Ioris
> http://unbit.it
> _______________________________________________
> uWSGI mailing list
> [email protected]
> http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
>
_______________________________________________
uWSGI mailing list
[email protected]
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
