httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sean Alderman <salderm...@udayton.edu>
Subject Re: [users@httpd] mod proxy balancer problem/question...
Date Tue, 02 Apr 2013 20:27:57 GMT
Perhaps I've not explained correctly, I apologize.  The concept I was going
for was to have two independent balancers that point to the same tomcat
instances.

In the webservices virtual host I would have a configuration like so:

  ProxyPass /service-sticky/ balancer://webservices-sticky/service-sticky/
  ProxyPassReverse /service-sticky/
balancer://webservice-sticky/service-sticky/

  ProxyPass /service-stateless/ balancer://webservices/service-stateless/
  ProxyPassReverse /service-stateless/
balancer://webservices/service-stateless/

In my config, both of the quoted Proxy balancer directives exist and are
referenced by a virtual host for different endpoints.  I absolutely
understand what the status=+H does, and I was hoping to use it to confine
all requests to services requiring stateful sessions to the tomcat instance
on tccontainer2 unless it is offline.

What I was trying to explain is that it would appear (by using mod status
to inspect requests sent to each balance member) that applying status=+H to
the balance member of one Proxy balancer, makes it apply to both, since the
address:port of the balance members are identical between the two
balancers.  I did not expect this behaviour and was looking for advice to
accomplish the goal of load balancing the stateless requests across both
containers and confining the stateful requests to one in a highly available
way.

Does that make more sense?


On Tue, Apr 2, 2013 at 4:06 PM, Igor Cicimov <icicimov@gmail.com> wrote:

>
> On 03/04/2013 2:02 AM, "Sean Alderman" <salderman1@udayton.edu> wrote:
> >
> > Greetings,
> >   I am running httpd 2.2.23.0-64 with mod_proxy to load balance Tomcat
> 6.0.36.B containers.  I have encountered a somewhat strange situation, and
> I was wondering if anyone could comment and or propose an alternative.
> >
> > I have a case where my tomcat containers have multiple webservice
> applications deployed.  Most of the deployments are stateless, but a few of
> them require session stickiness at the proxy layer.  I am looking for ways
> to better distribute the workload of the stateless webservice calls, with
> the hope of not having to create a new tomcat container separate stateful
> and stateless sessions.  The following configuration was tested, but had
> unexpected results...
> >
> > <Proxy balancer://webservices-sticky>
> >     BalancerMember ajp://tccontainer2.test.udayton.edu:12002route=webservices2-sticky
> >     BalancerMember ajp://tccontainer1.test.udayton.edu:12002route=webservices1-sticky
status=+H
> >     ProxySet lbmethod=byrequests
> >     ProxySet stickysession=JSESSIONID
> > </Proxy>
> >
> > <Proxy balancer://webservices>
> >     BalancerMember ajp://tccontainer1.test.udayton.edu:12002loadfactor=1 route=webservices1
> >     BalancerMember ajp://tccontainer2.test.udayton.edu:12002loadfactor=2 route=webservices2
> >     ProxySet lbmethod=byrequests
> > </Proxy>
> >
> > What I find is that balancer://webservices never sends any requests to
> ajp://tccontainer1.test.udayton.edu:12002.
>
> Thats because it never gets used, the requests are always being served by
> the first proxy. Why do you have 2 of them?
>
>  It would appear that the status=+H applies to the BalancerMember object
> instead of balancer://webservices-sticky.
>
> Correct, it means that that balancer member is hot standby as explained in
> the documentation.
>
>

Mime
View raw message