httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Vincent Janelle" <vjane...@home.com>
Subject Re: Multi-threaded apache rather than pre-forking
Date Tue, 26 Jan 1999 18:25:01 GMT
Sure, although have a funny feeling that it was all those nasty CGI's that
was at fault.

Some bad programming on our part.

I'm just not sure exactly what to check for on the side of apache..  But
I've seen it before, where the machine responds, but apache takes forever to
respond to connections.. I believe this has something to do with the
MaxClients and MaxRequestsPerChild directives that were setup there..

For some silly reason:

MaxClients 150
MaxRequestsPerChild 100

was

MaxClients 70
MaxRequestsPerChild 100.

Seems as though someone at varesearch was playing with "my perfect config".
Time to recompile apache and everything else and move to linux 2.2 =)

Just so you guys know, roxen was configured to run 25 threads and was
handing over 320requests/min, or 5 requests/second..  Not trying to troll or
anything, but thats what the load was like with the /. effect..  Its pretty
interesting to watch over 3-4 minutes and see the # of requests count past a
thousand when you fire up a webserver =)

>If you've already solved your problems using another web server, cool,
>congrats, perhaps apache isn't suited to your application.  You should
>read the perf-tuning documentation though
><http://www.apache.org/docs/misc/perf-tuning.html>.
>
>There is nothing we can do to deal with the "slashdot effect" -- no web
>server can pull magic rabbits out of the hat.  If you didn't test your
>site at that load then you can only blame yourself for its failure to
>handle that load.  Apache can certainly handle high load sites, you just
>need to engineer for them.
>
>But if you want to help us, you could give more details :)
>
>Dean
>


Mime
View raw message