[jacorb-developer] Error in Name Service / ORB

[Jörg Decker]

Hi Everyone,

we face a severe degrading issue with our production Jacorb setup. We
basically have a stable CORBA server running. But since we use the
notification service, we run into problems.

The ORB runs fine until we find the following exception in the ORB-Log:

2012-03-09 09:16:58.353 WARNING unexpected exception in runloop
    org.jacorb.util.threadpool.ThreadPool.putJob(ThreadPool.java:183)

org.jacorb.orb.giop.MessageReceptorPool.connectionCreated(MessageReceptorPool.java:84)
    org.jacorb.orb.BasicAdapter.add_input(BasicAdapter.java:402)
    org.jacorb.orb.etf.ListenerBase.deliverConnection(ListenerBase.java:129)

org.jacorb.orb.iiop.IIOPListener.deliverConnection(IIOPListener.java:309)
    org.jacorb.orb.iiop.IIOPListener.access$2200(IIOPListener.java:54)

org.jacorb.orb.iiop.IIOPListener$Acceptor.deliverConnection(IIOPListener.java:607)
    org.jacorb.orb.iiop.IIOPListener$Acceptor.run(IIOPListener.java:469)
2012-03-09 09:17:28.980 INFO Opened new server-side TCP/IP transport to
xx.xx.xx.xx:port
2012-03-09 09:17:28.980 WARNING unexpected exception in runloop
    org.jacorb.util.threadpool.ThreadPool.putJob(ThreadPool.java:183)

org.jacorb.orb.giop.MessageReceptorPool.connectionCreated(MessageReceptorPool.java:84)
    org.jacorb.orb.BasicAdapter.add_input(BasicAdapter.java:402)
    org.jacorb.orb.etf.ListenerBase.deliverConnection(ListenerBase.java:129)

org.jacorb.orb.iiop.IIOPListener.deliverConnection(IIOPListener.java:309)
    org.jacorb.orb.iiop.IIOPListener.access$2200(IIOPListener.java:54)

org.jacorb.orb.iiop.IIOPListener$Acceptor.deliverConnection(IIOPListener.java:607)
    org.jacorb.orb.iiop.IIOPListener$Acceptor.run(IIOPListener.java:469)

We noticed that the about 105 connections were opened to one IP that
belongs to the CORBA-Server (the one that serves the CORBA-Requests). Could
this be a resource leak or is this a normal situation?

2012-03-09 10:21:32.055 INFO Opened new server-side TCP/IP transport to
<server-ip>:port1
2012-03-09 10:21:32.055 INFO Opened new server-side TCP/IP transport to
<server-ip>:port2
2012-03-09 10:21:32.055 INFO Opened new server-side TCP/IP transport to
<server-ip>:port3
2012-03-09 10:21:32.055 INFO Opened new server-side TCP/IP transport to
<server-ip>:port4
... and so on ...

We are pretty much in the analysis phase of this issue. Since we did not
implement the connection handling, we might have an implementation issue
here. Basically, the notifications are delivered without noticable delays.
Though, as soon as the above exception occurs, the notification delay
increases until the whole system becomes totally unstable and needs to be
restarted.

Do you have any hints about this issue? Is there a known bug that could
have been fixed with version 3.0? Is there a parameter that we might set,
e.g. a connection idle timeout or max. amount of connections?

Your help is appreciated. Thanks,
Joerg