users@glassfish.java.net

Re: GFv3 + EclipseLink >=2.1.1: Remote EJB client cannot transfer JPA entities through remote session bean

From: Mitesh Meswani <mitesh.meswani_at_oracle.com>
Date: Thu, 16 Dec 2010 17:58:05 -0800

Did you ensure that same version of EclipseLink is used both on your
server and your appclient?

On 12/16/2010 5:34 PM, xavier_at_wflogistics.com wrote:
> Hi,
>
> I am having big troubles, I hope you can help me.
>
> I am using Glassfish v3.0.1, I needed to upgrade it to EclipseLink
> 2.1.1 because of a know bug (solved in 2.1.1) of EclipseLink 2.0.0 but
> this bring me a much bigger problem.
>
> I use Glassfish as a EJB server, a lot EJBs session beans are facades
> for the persistence layer (eclipselink), the remote java client connect
> to the server through remote EJBs to make CRUD. If I upgrade to
> EclipseLink 2.1.1 or newer when the remote java client connect to a
> remote session bean to send a JPA entity through a method in the remote
> session bean the connection hangs after a few minutes.
>
> The JPA entities classes are in a separated jar shared by the EAR in
> the server and the remote java client, the persistence.xml is in a EJB
> module inside a EAR in the server.
>
> This occurs with Glassfish 3.1 last nightly build, too. GF3.1 throws
> this:
>
> javax.ejb.EJBException: java.rmi.MarshalException: CORBA COMM_FAILURE
> 1330446344 Maybe; nested exception is:
> org.omg.CORBA.COMM_FAILURE: FINA: IOP00410008: Connection abort vmcid:
> OMG minor code: 8 completed: Maybe
> at
> com.sistemasaereos.model._ServiceBeanRemote_Wrapper.test(com/sistemasae
> reos/model/_ServiceBeanRemote_Wrapper.java)
> at enterpriseapplication2client.Main.main(Main.java:35)
>
> ...
> Caused by: org.omg.CORBA.COMM_FAILURE: FINA: IOP00410034: Received
> java.io.IOException: End-of-stream, in a blocking read on connection,
> SocketOrChannelConnectionImpl[
> java.nio.channels.SocketChannel[connected local=/192.168.60.107:49577
> remote=zeus.cdi.sistemasaereos.com.sv/192.168.60.10:55538] ESTABLISHED
> true true], because an end of stream was detected vmcid: OMG minor
> code: 34 completed: No
>
> ...
>
> If this is not caused by a new security mecanism in Eclipselink this
> should be a serious bug.
>
> For now I am stucked in Glassfish v3.0.1 with EclipseLink 2.0.0 trying
> to make a workaround with the know bug.