tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "David Rees" <dr...@greenhydrant.com>
Subject RE: mod_jk lbfactor strangeness
Date Fri, 10 Oct 2003 15:25:38 GMT
On Fri, October 10, 2003 at 5:47 am, tomcat@dormon.com sent the following
> On Fri, 10 Oct 2003, Ralph Einfeldt wrote:
>
>> 2. mod_jk works with sticky sessions so only new sessions
>>    are balanced. I belief but am not shure that it's just
>>    round robin.
>
> 2 Surely then the new session would be balanced to the faster machine,
> and I would see more activity on them.

You can use the manager app to count active sessions on each webapp.  You
will find that you may need to increase the lbfactor on certain instances
to get closer to an even load.  I have found that even new sessions do not
always get evenly distributed.  Rounding error somewhere perhaps?

>> 3. Bill Barker claims that the load balancing is broken
>>    as the instances of mod_jk don't know the load of each
>>    other. So mod_jk will balance to some extend but not as
>>    good as it could/should.
>
> 3 That would explain everything :(

The real issue is that you have some large number of processes and none of
them share their tomcat load info.  So with more than one Apache process,
things tend to not get distributed as expected.  If you use Apache2 with a
threaded MPM this is supposed to work better.

> Would a move to mod_jk2 be of any use, or should I get someone to put
> their hand in their pocket and upgrade the other two boxes

Better to use that $$$ and convince a Tomcat developer to fix mod_jk so
that it works properly.  ;-)

-Dave

---------------------------------------------------------------------
To unsubscribe, e-mail: tomcat-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: tomcat-user-help@jakarta.apache.org


Mime
View raw message