getting OIL error when using JVMServerILService

[cconnelly]

I am getting the following error under heavy load of my message queue:
java.net.PlainSocketImpl.socketAccept(PlainSocketImpl.java:Native method)
java.net.PlainSocketImpl.accept(PlainSocketImpl.java:343)
java.net.ServerSocket.implAccept(ServerSocket.java:438)
java.net.ServerSocket.accept(ServerSocket.java:409)
org.jboss.mq.il.oil.OILClientILService.run(OILClientILService.java:127)
java.lang.Thread.run(Thread.java:536)


What I do not understand is that I am using the default configuration (JBoss 3.05), which says that I am using org.jboss.mq.il.jvm.JVMServerILService as my JMS invocation layer for persistance.

If I am truly using the JVM invocation layer, why is JBoss opening up sockets, and why is it using the org.jboss.mq.il.oil.* library?

I can only guess that jboss either sharing the oil API behind the scense for the JVM invovocation layer, or there is some sort of bug in jboss. The XAConnectionFactoryJNDIRef is very similar for OIL and JVM, with OIL being "XAConnectionFactory" and JVM being "java:/XAConnectionFactory".

[adrian.brock]

Show the code you use to get the connection

Regards,
Adrian

[cconnelly]

public Object invoke(Object[] args)
throws NamingException, JMSException, Throwable {

log.debug("JMSInvocationConnector.invoke()");
// initialize MethodInvocation
String methodName = "invoke";
MethodInvocation mi = new MethodInvocation();
mi.setMethodName(methodName);
mi.setParams(args);
ObjectMessage msg = null;

int ack = Session.AUTO_ACKNOWLEDGE;

log.debug("JMSInvocationConnector.constructor:queue");

QueueConnectionFactory factory = null;
try {
ctx = new InitialContext();
factory = (QueueConnectionFactory) ctx.lookup("ConnectionFactory");
queue = (Queue) ctx.lookup("queue/primaryMBeanQueue");
} catch (NamingException e) {
log.error("JMSInvocationConnector.constructor: naming exception");
throw e;
}

log.debug("JMSInvocationConnector.looking up queue and send");

try {
connect = factory.createQueueConnection();
queueSession = connect.createQueueSession(false, ack);
sender = queueSession.createSender(queue);
connect.start();

sender = queueSession.createSender(queue);
msg = queueSession.createObjectMessage(mi);
sender.send(msg);
sender.close();
} catch (JMSException e) {
log.error("JMSInvocationConnector.JMSException:" + e.getMessage());
} finally {
if (connect != null) {
try {
connect.close();
} catch (JMSException e) {
log.error("JMSInvocationConnector.JMSException:" + e.getMessage());
}
}
}
return null;
}

[cconnelly]

Figured it out, I was using the wrong JNDI name to connect to the ConnectionFactory. Thanks for leading me in the right direction, I thought that the Invocation Layer was purely a JMS configuration issue.

[adrian.brock]

For others it should be

factory = (QueueConnectionFactory) ctx.lookup("java:/ConnectionFactory");

Even better is
factory = (QueueConnectionFactory) ctx.lookup("java:/JmsXA");
since that uses JCA to link into the EJB transaction.

Regards,
Adrian

[cconnelly]

"java:/JmsXA" does this help avoid potential deadlock when the queue listener accesses entity beans.

[adrian.brock]

? Not sure what this has to do with entity
deadlocks.

A deadlock is due to access in different orders.

Thread1
lock A

Thread2
lock B

Thread1
try to lock B and wait

Thread2
try to lock A and wait

Both will timeout unless the deadlock is detected
and broken earlier.

Regards,
Adrian