If "Request Meaaage" is more than the number of threads in "thread-pool-1",all pooled threads become time-outs.

[glassfishrobot]

If "Request Meaaage" that is more than the number of threads in "thread-pool-1" is sent to the IJServer cluster,
all pooled threads wait for "Fragment Message" to be processed and become time-outs.

**LOG**javax.ejb.EJBException: java.rmi.MarshalException: CORBA COMM_FAILURE 1330446361 No; nested exception is: 
	org.omg.CORBA.COMM_FAILURE: WARNING: 00410025: Write of message exceeded TCP timeout : max wait time = 6,000 ms, total time spent blocked, waiting to write = 7,280 ms.  vmcid: OMG  minor code: 25  completed: No
javax.ejb.EJBException: java.rmi.MarshalException: CORBA COMM_FAILURE 1330446361 No; nested exception is: 
	org.omg.CORBA.COMM_FAILURE: WARNING: 00410025: Write of message exceeded TCP timeout : max wait time = 6,000 ms, total time spent blocked, waiting to write = 7,280 ms.  vmcid: OMG  minor code: 25  completed: No
	at com.fujitsu.test.ejb._DataTest_Wrapper.sendData(com/fujitsu/test/ejb/_DataTest_Wrapper.java)
	at datatestacc.InnerThread_loop1.run(Main.java:66)
Caused by: java.rmi.MarshalException: CORBA COMM_FAILURE 1330446361 No; nested exception is: 
	org.omg.CORBA.COMM_FAILURE: WARNING: 00410025: Write of message exceeded TCP timeout : max wait time = 6,000 ms, total time spent blocked, waiting to write = 7,280 ms.  vmcid: OMG  minor code: 25  completed: No
	at com.sun.corba.ee.impl.javax.rmi.CORBA.Util.mapSystemException(Util.java:258)
	at com.sun.corba.ee.impl.presentation.rmi.StubInvocationHandlerImpl.privateInvoke(StubInvocationHandlerImpl.java:211)
	at com.sun.corba.ee.impl.presentation.rmi.StubInvocationHandlerImpl.invoke(StubInvocationHandlerImpl.java:150)
	at com.sun.corba.ee.impl.presentation.rmi.codegen.CodegenStubBase.invoke(CodegenStubBase.java:226)
	at com.fujitsu.test.ejb.__DataTest_Remote_DynamicStub.sendData(com/fujitsu/test/ejb/__DataTest_Remote_DynamicStub.java)
	... 2 more
Caused by: org.omg.CORBA.COMM_FAILURE: WARNING: 00410025: Write of message exceeded TCP timeout : max wait time = 6,000 ms, total time spent blocked, waiting to write = 7,280 ms.  vmcid: OMG  minor code: 25  completed: No
	at com.sun.proxy.$Proxy36.transportWriteTimeoutExceeded(Unknown Source)
	at com.sun.corba.ee.impl.transport.NioBufferWriter.write(NioBufferWriter.java:62)
	at com.sun.corba.ee.impl.transport.ConnectionImpl.writeUsingNio(ConnectionImpl.java:508)
	at com.sun.corba.ee.impl.transport.ConnectionImpl.write(ConnectionImpl.java:480)
	at com.sun.corba.ee.impl.encoding.CDROutputObject.writeTo(CDROutputObject.java:225)
	at com.sun.corba.ee.impl.transport.ConnectionImpl.sendHelper(ConnectionImpl.java:1153)
	at com.sun.corba.ee.impl.transport.ConnectionImpl.sendCancelRequest(ConnectionImpl.java:1142)
	at com.sun.corba.ee.impl.transport.ConnectionImpl.sendCancelRequestWithLock(ConnectionImpl.java:1162)
	at com.sun.corba.ee.impl.protocol.MessageMediatorImpl.sendCancelRequestIfFinalFragmentNotSent(MessageMediatorImpl.java:373)
	at com.sun.corba.ee.impl.protocol.ClientRequestDispatcherImpl.endRequest(ClientRequestDispatcherImpl.java:916)
	at com.sun.corba.ee.impl.protocol.ClientDelegateImpl.releaseReply(ClientDelegateImpl.java:279)
	at com.sun.corba.ee.impl.presentation.rmi.StubInvocationHandlerImpl.privateInvoke(StubInvocationHandlerImpl.java:208)
	... 5 more

To operate even in this case, I propose the following corrections.

(1) The thread pool for the thread that processes "Fragment Message" is prepared.
ex) "fragment-thread-pool"

(2) Received "Fragment Message" is processed by the thread pooled by (1).

**glassfish-corba/orbmain/src/main/java/com/sun/corba/ee/impl/protocol/giopmsgheaders/MessageBase.java**
    public static MessageBase parseGiopHeader(ORB orb,
              Connection connection,
              ByteBuffer buf,
              int startPosition) {
        ...

        //
        // Initialize the generic GIOP header instance variables.
        // 
        if ((it[4] == 0x01) && (it[5] == 0x00)) { // 1.0
            Message_1_0 msg10 = (Message_1_0) msg;
            msg10.magic = magic;
            msg10.GIOP_version = new GIOPVersion(it[4], it[5]);
            msg10.byte_order = (it[6] == LITTLE_ENDIAN_BIT);
            // 'request partitioning' not supported on GIOP version 1.0
            // so just use the default thread pool, 0.
            msg.threadPoolToUse = 0;
            msg10.message_type = it[7];
            msg10.message_size = readSize(it[8], it[9], it[10], it[11],
    msg10.isLittleEndian()) +
    GIOPMessageHeaderLength;
        } else { // 1.1 & 1.2
            Message_1_1 msg11 = (Message_1_1) msg;
            msg11.magic = magic;
            msg11.GIOP_version = new GIOPVersion(it[4], it[5]);
            msg11.flags = (byte) (it[6] & TRAILING_TWO_BIT_BYTE_MASK);
            // IMPORTANT: For 'request partitioning', the thread pool to use
            //            information is stored in the leading 6 bits of byte 6.
            //
            // IMPORTANT: Request partitioning is a PROPRIETARY EXTENSION !!!
            //
            // NOTE: Bitwise operators will promote a byte to an int before 
            //       performing a bitwise operation and bytes, ints, longs, etc
            //       are signed types in Java. Thus, the need for the 
            //       THREAD_POOL_TO_USE_MASK operation.
            if(it[7] == GIOPFragment){           // add
msg.threadPoolToUse = orb.getThreadPoolManager().getThreadPoolNumericId("fragment-thread-pool"); // add
            } else{
msg.threadPoolToUse = (it[6] >>> 2) & THREAD_POOL_TO_USE_MASK;
            }
            msg11.message_type = it[7];

        ...

We ask for your kind consideration.

[glassfishrobot]

Reported by myokan

[glassfishrobot]

myokan said:
It does not become time-outs. It happened because of the proposed correction.

[glassfishrobot]

This issue was imported from java.net JIRA GLASSFISH_CORBA-17

[glassfishrobot]

Closing this as this issue is migrated to eclipse-ee4j/orb#17