Problem: Product Name: Borland Application Server Product Version: 6.5, 6.6 and 6.7 Product Component: EJB Container/JMS Platform/OS version: All JDK: All Patch Level: All If an application has defined the waitTimeout property in the Borland EJB deployment descriptor for the Message Driven Bean (MDB) and there are no more MDB in the pool, the Borland Application Server will throw a JMSException to the Sonics JMS service provider to wait if the MDB has exceeded the defined time period. The JMSException has caused the Sonics thread for the ConnectionConsumer to die and stop pushing JMS message to the Borland Application Server. Resolution: JMS Provider and Application Server Relationship The Application Server implements the ServerSession and ServerSessionPool. It wills also control the thread for the session. The JMS Provider implements the ConnectionConsumer and Session. The application implements the MessageListener interface for the Message Driven Bean. When a JMS message arrives at the JMS provider queue, it will create a ConnectionConsumer object. In the case of Sonics, it relies on a thread to execute the ConnectionConsumer object. The ConnectionConsumer will attempt to get a ServerSession from the ServerSessionPool. If there is no more ServerSession, the ConnectionConsumer will wait. If a ServerSession is available, the ConnectionConsumer will use a reference of the ServerSession to create the JMS session object. The ConnectionConsumer will deliver the message to the Session and the application server will deliver the JMS message to the MessageListener implemented at the Message Driven Bean. If there is no more ServerSession in the pool, the ConnectionConsumer will wait. The workaround to prevent the Borland Application Server from stopping to process the JMS messages is to set the waitTimeout property to 0. References Java(TM) Message Service Specification Final Release 1.1
↧