Hi,
We are seeing a strange phenomenon where on a clustered glassfish V2 deployment unused EJBs are initialized (actually it is a sailfin deployment but SIP is not used).
In particular an application is deployed that provides a remote EJB but which is not used in the test scenario. The initial pool size of the EJB pool is 200 and we are seeing these beans being initialized some time after startup. The strange thing is that this happens rather late after startup and at the same time on all application server instances (in our example 10 minutes 12 seconds).
What is being tested in this case is a resource adapter which is handling traffic and using the work manager of JCA to dispatch this traffic to an application, where the application server is restarted under load. There is also one thread pool in effect in the application server.
Therefore, one might assume that the thread pool is utilized fully by the resource adapter at startup and that events to initialize EJB pools of unused applications are dispatched later. However, this does not explain why the initialization of unused EJBs takes place at the same time cluster wide.
Does anyone have any idea why this behavior is there? We just want to understand what is happening.
Cheers
Erik
[Message sent by forum member 'erikengerd' (erik_at_brakkee.org)]
http://forums.java.net/jive/thread.jspa?messageID=388431