Apache Plugin is not working with Cluster Entry in httpd.conf

Question

Hi Guys,
We have a very particular issue that is happening here in our production environments. Before explaining the issue to you, I will just give a brief of our architecture.
We are using Weblogic Portal Platform SP4.
Our current Production Domain Consists of 1 Admin Server and 2 Managed Servers. The managed servers are in a cluster.
We have two Linux boxes on which the weblogic servers are running. These two machines have IPs: x.x.x.58 and x.x.x.59. Now, the Admin Server and managed Server 1 are started on the machine x.x.x.58 and managed server 2 is started on the machine x.x.x.59. Both the managedServers listen on port 8007.
This entire setup mentioned in the above paragraph is at a location A (lets say).
Now, the requests are proxied to the managed server cluster using BEA's Apache plugin residing on Apache. Apache is running on two Linux boxes that are present at a location B.
Location B is connected to Location A by a dedicated WAN Link.
Now, before the two Apache servers there is a physical Load Balancer that balances the requests between the two Apache boxes.
This means that whenever there is a request from the Internet to the weblogic cluster, the load balancer first sends it to either of the Apache servers and then the BEA Apache plugin on that Apache forwards the request via the 2 mbps WAN link to the weblogic cluster.
Now following is the description of the issue we are facing. We donot know whether the issue is with the plugin or with weblogic server.
In order that the plugin load balances the requests between the servers in the cluster and for failover, the following entry was put in httpd.conf on both the Apache Servers.
WebLogicCluster y.y.y.11:8007,y.y.y.13:8007
Note: Because of the WAN link there is natting used. This implies y.y.y.11 is natted to x.x.x.58 and y.y.y.13 is natted to x.x.x.59.
Here in lies the issue. When we use this plugin entry the following issue occurs.
1.     First we go to the home page and login.
2.     The user is then redirected to a Portal Desktop fitting those credentials.
3.     Now on this Portal Desktop, invariably whenever we click any of the desktop's links, an HTTP 403 is thrown. This error is thrown because on the desktop there are entitlements so that it cannot be accessed without a user logging in with credentials. But this error indicates that the server is losing the session and entitlement details.
Why does this happen?
As a result, we cannot use that plugin entry. The current entry we have is
WebLogicHost y.y.y.11
WebLogicPort 8007
As you can see, here we are only redirecting to one managed server. This entry is not what we want, obviously because if that server fails, the plugin wont know which other server to pass the request on to. Also in this case there is no load balancing.
Our issues our,
1. How to resolve this plugin entry issue. How can we use the cluster entry without encountering this HTTP 403 Error?
3. Consider the following sequence of steps.
- Request for the home page is served by Apache Server 1
- User logins in, the login request is sent by the Load Balancer to Apache Server 1 and plugin on Apache Server 1 sends it to the cluster via WAN l ink. Then the internal mechanism works with creating the session id and primary server and secondary server and logs the user in and sends the response via Apache Server 1.
- Now, the user clicks a link on the desktop. The request now goes to the load balancer and it sends it to Apache Server 2.
The question is now does the plugin on Apache server 2 correctly understand and send the request to the primary server of that session.
Pls help me as to giving me some direction in resolving this issue

Answer 1

From your Xorg.0.log:
(II) fglrx(0): DRIScreenInit done
(II) fglrx(0): Kernel Module Version Information:
(II) fglrx(0): Name: fglrx
(II) fglrx(0): Version: 8.49.7
(II) fglrx(0): Date: May 12 2008
(II) fglrx(0): Desc: ATI FireGL DRM kernel module
(WW) fglrx(0): Kernel Module version does *not* match driver.
(EE) fglrx(0): incompatible kernel module detected - HW accelerated OpenGL will not work
You have to quit X, then rmmod and modprobe the kernel module - you're still using the old one.  (a reboot would technically fix it too)
# rmmod fglrx
# modprobe fglrx
# dmesg | tail
The dmesg output should say you've successfully loaded the 8.501 driver.