Difficulties with _clone() calls

[Jerome Louvel]

Hi !

We are developing a Visibroker for Java multi-threaded application
using exactly one IIOP connection per thread for accessing the same
implementation object on the same OrbixWeb server.

For this we use the clone() method which effectively create a new TCP
socket for each thread.

However, a problem appear when we try to connect simultaneously several
thread (50). Some threads successfully connect but rapidly the following
exception is raised :

org.omg.CORBA.COMM_FAILURE[completed=MAYBE,
reason=java.net.ConnectException: Connection refused]
at com.visigenic.vbroker.orb.TcpSocket.clone(Compiled Code)
at com.visigenic.vbroker.orb.TcpConnection.clone(Compiled Code)
at com.visigenic.vbroker.orb.GiopConnectionImpl.clone(Compiled
Code)
at
com.visigenic.vbroker.orb.GiopStubDelegateServerInfo.clone(Compiled
Code)
at com.visigenic.vbroker.orb.GiopStubDelegate.clone(Compiled
Code)
at com.visigenic.vbroker.orb.UnboundStubDelegate.clone(Compiled
Code)
at org.omg.CORBA.portable.ObjectImpl._clone(Compiled Code)
at
at benchbuilder.injector.client.Client.run(Compiled Code)

We suppose that the problem comes from the server because when we
replace
the Orbix server with a Visibroker equivalent (same source, same
computer)
the connections works fully.

Is our supposition right ? In this case, are there some configuration
adjustments on the Orbix server or on the Visibroker client like TCP
time-outs or connection retries that we should try ?

Thanks for helping,

Nicolas Jean, Jerome Louvel (ASTON)

[Boris Goldberg]

Jerome Louvel wrote:
>
> Hi !
>
< We are developing a Visibroker for Java multi-threaded application
< using exactly one IIOP connection per thread for accessing the same
< implementation object on the same OrbixWeb server.

< For this we use the clone() method which effectively create a new TCP
< socket for each thread.

< However, a problem appear when we try to connect simultaneously
several
< thread (50). Some threads successfully connect but rapidly the
following
< exception is raised :
<
< org.omg.CORBA.COMM_FAILURE[completed=MAYBE,
< reason=java.net.ConnectException: Connection refused]

> at org.omg.CORBA.portable.ObjectImpl._clone(Compiled Code)

> at
> at benchbuilder.injector.client.Client.run(Compiled Code)
>
< We suppose that the problem comes from the server because when we
< replace
< the Orbix server with a Visibroker equivalent (same source, same
< computer)
< the connections works fully.

< Is our supposition right ? In this case, are there some configuration
< adjustments on the Orbix server or on the Visibroker client like TCP
> time-outs or connection retries that we should try ?

Correct. The problem is server's: its TCP listen is too small. I don't
know if it's configurable (I don't have the docs). If it's not, have
you threads connect after a brief interval.