Weblogic Stateless Session methods invoked from different cluster member servers - ejb

Question

My environment is Weblogic 10.3.5 on Solaris box. EJB is version 3 and there is anotation in the Bean class. Sorry for the confusion as the code is new to me and they also have deployment descriptor to generate ejb2 client code for another client to call, so it's not straigtforward.
I have a stateless session bean deployed to a cluster which has 2 server members say they are member1 and member2.
The session bean is deployed as clusterable as this is in the anotation:
homeIsClusterable = Constants.Bool.TRUE
This is how my Stand alone Java client lookup and call the EJB methods:
private void testBean(){
bean.methodA();
bean.methodB();
}
In the provider URL I ONLY specify the provider URL to ONE server member:
env.put(Context.PROVIDER_URL, "t3://member1:7005");
env.lookup("remote#the.bean.qulified.remoteinterface")
The Jndi name above is using the "mapped name + qualified remote interface class name", the mapped name is defined in the anotation.
Now the problem is, I found out, bean.methodA() got invoked in member1, and methodB() got invoked on member2, I found this from the logs of each server member. So it's always like this, member1 log will only show debug information from methodA, and member2 will only show debug information from methodB.
So here is my conceptual question - is this possible at all ? Are the above 2 methods supposed to be called on member1 only ? I know it's possible when you lookup through home interface you could possibly get a bean from either server, but in this case the ejb3 lookup is not going through the home interface(like in ejb2 we get a home and then call create method) but directly getting a remote object.
This caused issue as our methodB has a dependancy on methodA(methodA is doing some cleanup job, and then method re-initialize the cache), we need to do this on each cluster member.
This is just extra info but please focus on the above question from a concept perspective.

Answer 1

From the documentation:
When home-is-clusterable is True, the EJB can be deployed from multiple WebLogic Servers in a cluster. Calls to the home stub are load-balanced between the servers on which this bean is deployed, and if a server hosting the bean is unreachable, the call automatically fails over to another server hosting the bean.
I believe this is the case even when you explicitly only connect to a single member. This has some pretty good info in the Replica-Aware Home section:
http://www.informit.com/articles/article.aspx?p=101737&seqNum=8
It's more or less the whole point of clustering... a cluster appears as if it's a single server instance to a client.