-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Martin,

On 3/22/12 5:08 PM, Martin Registe, Jr. wrote:
>> First question: do all httpd instances know about all Tomcat
>> instances? Or, does httpd only know about the two instances that
>> are co-located with it on the same physical server?
> 
> Httpd only know about the two instances that re collocated with it
> on the same physical server

Okay.

>> If https only knows about its local 2 Tomcat instances, then
>> you'll only need 2 workers, etc. I have no idea how this could
>> possibly work with sticky sessions unless the hw lb knows how to
>> dispatch to the right http/2xTomcat physical server, in which
>> case you are wasting your time with all these extra processes.
> 
> The hw loadbalancer sends the traffic to 10.17.75.61 port 80... 
> apache takes it and using the workers.properties file sends it to
> one of TC instances then apache sends it back through the hw lb it
> has a sessionid and jvm tag on in the host header so it knows which
> jvm it came from

So, when a client first arrives, they are randomly-assigned an httpd
instance and then randomly-assigned a Tomcat instance that "belongs"
to that httpd instance. The hw lb later uses the "Host" header to
determine which httpd instance gets subsequent requests?

The "Host" header should include the IP address or hostname of the
target server. Are you telling the client that the Host changes when
they are assigned a Tomcat server? I still don't understand why you
don't just have a more-connected cluster where all httpds are
connected directly to all Tomcats.

>> What about <Engine>? It would be nice to confirm that your
>> jvmRoute(s) is(are) set correctly.
> 
> TC1    <Engine name="Catalina" defaultHost="localhost"
> jvmRoute="jvm1"> TC2    <Engine name="Catalina"
> defaultHost="localhost" jvmRoute="jvm2">

Looks good.

> My next thoughts I don't know exactly how this is working but it
> definitely is working and doesn't seem to be getting sessions
> jumping from one jvm to the other (on the same physical server).

Uh... okay. I thought you were coming here for help. Do you not
actually have a problem? If your boss simply suspects that there is a
problem, make him prove it via observations. You can't prove a
negative so it's his job to show you a problem, not your job to prove
that no problem exists.

> I will play around with the configuration some more but this has
> been working for over a year. You are saying if I just go with
> 
> worker.jvm1.type=ajp13 worker.jvm1.host=localhost 
> worker.jvm1.port=8019 worker.jvm1.lbfactor=10
> 
> worker.jvm2.type=ajp13 worker.jvm2.host=localhost 
> worker.jvm2.port=8029 worker.jvm2.lbfactor=10
> 
> worker.part.type=lb worker.part.balance_workers=jvm1,jvm2 
> worker.part.sticky_session=1
> 
> everything will continue to work, correct?

It should: you are simply balancing between two Tomcat instances and
using sticky sessions. It's about as simple a configuration as you can
have.

- -chris
-----BEGIN PGP SIGNATURE-----
Version: GnuPG/MacGPG2 v2.0.17 (Darwin)
Comment: GPGTools - http://gpgtools.org
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk9rneAACgkQ9CaO5/Lv0PA2GACeP3uB1M6znUXhcZO+FWxk5I8Y
tJ0AoK6MLicNEz/68bsJs0ZB50xJ9GGf
=pndJ
-----END PGP SIGNATURE-----

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org

Reply via email to