I believe it has been explained several times that replication of stateful services is not supported. Yet! :-)
My question, though, is the following:
which way would you recommend going if my setup is a clustered JBossAS5.1.0 + JBossESB4.6 (jUDDI, JBM clustered) and my requirement is to have failover support of services? (in my case the collection of services to be deployed in the cluster would comprise an Aggregator)
I mean, I do not have a specific requirement whereby I need to be processing all those splitted messages coming into the Aggregator in a distributed manner but... looking at the Aggregator code I was wondering if using a distributed cache on the Map used to store the splits would not do the trick and allow me to forget about failover support since all splits are coming in through a clustered JMS queue...
Right... as it stands, the Aggregator will not work in a clustered fashion. Try modifying it to use a distributed cache and let us know how you go :)