strange behaivor of nonclustered ejb calling clustered ejb

[wurzelkuchen]

Hello,

during my work on a student research project, I noticed a strage behavior.
Following scenario:

EJB1: a not clustered stateless EJB, which calls EJB2 per lookup

@Stateless
public class Bean implements BeanLocal, BeanRemote {
 ...
 private Bean2 bean2;
 public void call(){
 Properties properties = new Properties();
 properties.put("java.naming.factory.initial",
 "org.jnp.interfaces.NamingContextFactory");
 properties.put("java.naming.factory.url.pkgs",
 "org.jboss.naming:org.jnp.interfaces");
 properties.put("java.naming.provider.url",
 "192.168.0.52:1100");
 Context context;
 try {
 context = new InitialContext(properties);
 bean2 = (RemoteBean2) context.lookup(Bean2.RemoteJNDIName);
 while(...){
 bean2.doSomething();
 }
 } catch (NamingException e) {
 e.printStackTrace();
 }
 }

EJB2: a simple clustered stateless EJB

@Clustered
@Stateless
public class Bean2 implements BeanLocal2, BeanRemote2 {
 ...
 public void doSomwthing(){
 ...
 }
}

So, if I start these Beans like above, Bean2 starts only on one node of my Cluster. If I use port 1099 (normal JNDI instead of HAJNDI), Bean2 starts on different nodes each time I start bean2.doSomething().
I realy don't undestand this behaivior. Why does the use of HA-JNDI lets the bean2 unclustered while the use of JNDI leads to a clustered behaivior? I would understand if HAJNDI leads to a clustered and JNDI to an unclustered behaivior, but the opposite happens.
Does anybody have got a good explanation or at least an idea why?

Best Regards