Second and folowing nodes take very long time to come up

[sheckler]

I tried to set up clustering of an application which consists of about 50 SLSBs deployed as one Ear file. The Sessions Beans are declared as clustered and use a container configuration with a clustered invoker.

Both nodes are running JBoss 3.2.8.SP1 on W2K.

1st behaviour:

Cluster behaviour depends on th start order:

a) First node A, then node B -> 1 Cluster with 2 members
b) First node B then node A -> 2 Single server

2nd behaviour:

For case a) the deployment of the Earfile with the clustered SLSBs of the second node takes about 1 hour (on the first the whole startup takes about 1 minute!), but then the server works and a connected client can lookup the SLSBs and Load Balancing works.

An additional problem is, that a connected client already contacts server B getting an exception (EJB not bound) within that time.

I tried to start with run -c ABC -b pc-000[1|2] as well as run -c ABC with same behaviour.

The configuration used is based on the "all" configuration.


Am I struggeling with network issues or misconfiguration problems?
Any help is appreciated
Thanks

[sheckler]

I addition I tested running 2 instances on 1 machine with binding manager.
The behaviour does not change. The deployment of clustered SLSBs takes very long time on the 2nd instance, while the logfiles (JGroups trace, jboss debug) show no errors or warnings. So network issues are not the reason.

The time is taken after the SecurityProxy initialization messages. What happens at that point?

When I stop the first instance in this phase, the startup of the second instance returns to "normal speed".

Please can someone give me a hint, what is happening at this point behind the scenes. Has someone seen similar effects?

Thanks

[sheckler]

I figured it out. The reason for that behaviour was a system property "java.naming.provider.url" (node1:100,node2:1100) set on the server.