httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mike Franon <kongfra...@gmail.com>
Subject Re: [users@httpd] apache 2.4 with MPM Event recommend config
Date Mon, 29 Apr 2013 17:07:05 GMT
Thanks for the info so for 2.4 it would be very similar?

For us we get anywhere form 10 RPS to 20 RPS on average and during peak
periods or sales/promos we get upwards of 150-200 RPS

Thanks


On Fri, Apr 26, 2013 at 11:50 AM, Tom Evans <tevans.uk@googlemail.com>wrote:

> On Fri, Apr 26, 2013 at 3:41 PM, Mike Franon <kongfranon@gmail.com> wrote:
> > Hi,
> >
> > I recently moved to apache 2.4.4 with php-fpm, mod_fastcgi, and MPM
> Event.
> >
> > I was curious what values do I need to change to optimize apache?
> >
> > I remember in 2.4 prefork I had to change several like MaxClients and
> > MaxRequestsPerChild
> >
> >
> >
> > The server I am running it on has 30 GB Ram, and 8 Cores
> >
> >
> > I know usually the defaults are not great out of the box and curious if
> you
> > have any recommendations to get the best performance possible for apache
> 2.4
> >
> > Thanks
> >
> >
>
> What are your average and max concurrent requests per second that you
> want to achieve? Simply tune to that - you will find apache with event
> MPM (and without language interpreters embedded in httpd, eg using
> php-fcgi) does not use much memory or CPU at all.
>
> Eg, for our moderately loaded proxies, which use apache 2.2 with event
> MPM, we have average of around 50 RPS (per proxy, there are a few of
> them), with peaks of around 500 RPS. We configure the server:
>
>     StartServers 8
>     MaxClients 1024
>     MinSpareThreads 128
>     MaxSpareThreads 768
>     ThreadsPerChild 64
>     MaxRequestsPerChild 0
>
> So we start 8 servers, each with 64 threads, giving us 512 request
> slots immediately. Whenever there are less than 128 spare slots, we
> start an extra server, up to a maximum of 1024 slots (16 servers), and
> we reap a server whenever there are more than 768 spare slots. We also
> trust that Apache without interpreters does not leak RAM, so don't
> bother reaping servers after a certain number of handled requests.
>
> Our normal and peak usage fits in to the slots started initially, with
> scope to grow if needed. These servers only serve static files off
> disk, or proxy requests to backend servers. Memory usage and load
> remain steady on each proxy, apache uses less than 1 GB of RAM, load
> average has never been above 1.
>
> Hope that helps
>
> Cheers
>
> Tom
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
> For additional commands, e-mail: users-help@httpd.apache.org
>
>

Mime
View raw message