tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mladen Turk <mladen.t...@jboss.com>
Subject Re: mod_jk connection_pool_size property
Date Tue, 01 Aug 2006 14:26:11 GMT
Java Rab wrote:
> I have setup load balancing between Apache 1.3.33 and Tomcat 4.0.6 using
> the mod_jk connector. Initial tests were fine and the balancing seemed to
> work a treat. However, with any load on the server, I was seeing 503 errors
> reporting server unavailability. To resolve this issue, I set the following
> property in the workers.properties file: connection_pool_size=1000 which 
> has
> resolved the immediate problem and the system can now process a heavy load
> without failure.
> 
> The only thing that troubles me, that's preventing me from claiming success
> is the following note in red highlighting in the apache documentation, 
> which
> can be viewed online at
> http://tomcat.apache.org/connectors-doc/config/workers.html. Extract: "Do
> not use connection_pool_size with values higher then 1 on *Apache 
> 2.xprefork
> * or *Apache 1.3.x*!". Can anyone explain the reason why this shouldnt be
> done? My system seems to work fine, but I'd be nervous going live with the
> configuration described above, without first understanding the warning in
> the apache docs.
> 
> Any help will be great!
> 

Apache 1.3 and Apache 2.x-prefork create a separate child process for
each client connection, meaning that if you have default
250 MaxClients, you can end up with MaxClients * connection_pool_size
connections to the Tomcat, that would in your case be 250000.
Pretty high number, right :)

With worker mpm the connection_pool_size defaults to ThreadsPerChild,
that in the final gives the MaxClients connections to the Tomcat.
So the default value for connection_pool_size is always ThreadsPerChild
that with prefork mpm is by design always 1.

I suggest that you Google for worker/prefork mpm theory and usability.
Lots of OS vendors simply choose prefork mpm without any analysis,
simply presuming its more safe then any semi threaded model.

Regards,
Mladen.

---------------------------------------------------------------------
To start a new topic, e-mail: users@tomcat.apache.org
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org


Mime
View raw message