From: Dick Davies <rasputnik_at_hellooperator.net>
Date: Tue, 4 Nov 2008 12:46:14 +0000

I've had a good dig around in the meantime and have found

* GMS is enabled
* multicast is setup in the OS (can ping the multicast IP ok)
* sun-web.xml in the warfile is set to:

  * persistence type : replicated
  * persistence frequency : web-method
  * persistence scope : session

and the cluster-config looks like the attached screenshot.

That *is* in-memory replication, right? So the fact that HTTP Session
Store points to 'jdbc/hastore' (which doesn't exist) is OK?

And should I hane 'Persistence Store Health Check' enabled?

Thanks for any pointers you can give.




On Mon, Nov 3, 2008 at 3:27 PM, Dick Davies <rasputnik_at_hellooperator.net> wrote:
> Hi all
>
> I've got a 2 node cluster
> (runnning the Roller weblog engine) that doesn't seem to be
> replicating it's sessions correctly.
>
> If I set the load balancer to only send to nodeA, log in, then flip
> the backends so only nodeB is in
> then I have to log in again.
>
> Does anyone know if there's a checklist of what needs to be configured
> so HTTP sessions will
> replicate correctly?
>
> Also, is there any instrumentation or other way to look at the
> instances session caches. That would
> be really really useful to troubleshoot this.
>
> Thanks a lot.
>