I deployed a set of entity beans in my jboss as clustered beans. I have also a stateless session bean which handles those entity beans (local interfaces). I was wondering what information is really clustered for entity beans in cluster (in real, my entity beans are flagged as cluster but they extends the Standard CMP 2.x EntityBean with cache invalidation rather than Clustered CMP 2.x EntityBeanbecause I need cache invalidation...), since entity information cannot be replicated since I have only local interfaces (and I don't know if it makes sense since data is database).
I could be doing a big confusion here but if information is not passed by value from one instance to the other one (local interfaces), is jboss passing any kind of information to the another jboss instances to prepare/load those referenced entity beans instances, in case of future requests and load-balancing issues?
If I am not mistaken, I read something about only homes are prepared on cluster. Is this correct? What is the real advantage of doing it against the non-clustered beans? I mean, in a jboss instance I only can access the entity beans in that instance since I'm using only local interfaces. So, the current jboss instance never requests any information from another instances (weight load-balancing issues), unless for locality/time cache references.