Configure Hibernate Search/ActiveMQ for Amazon EC2

dungleonhart · **Joined:** Sat Jul 23, 2011 12:28 am **Posts:** 8

Hi guys,

I'm working on a project that is deployed to Tomcat servers on Amazon EC2 and used Hibernate Search. I'm trying to follow the instructions in this article: http://community.jboss.org/wiki/ClusteringUsingActiveMQAndTomcat, but I've got stuck at the getSession() method of the AbstractJMSHibernateSearchController class.

Here's the class that extends the AbstractJMSHibernateSearchController class in my master application:

Code:

public class HSearchListener extends AbstractJMSHibernateSearchController {      
    // My problem is here
    @Autowired      
    private SessionFactory sessionFactory;            
 
    @Override      
    protected Session getSession() { 
        return sessionFactory.getCurrentSession();      
    }       
    ///

    @Override      
    protected void cleanSessionIfNeeded(Session arg0) { } 

    public void register() { 
        // Copied from the referenced article          
        try {                
            Context initCtx = new InitialContext();                 
            // Look up for connection factory & create connection, session
            ConnectionFactory connectionFactory = (ConnectionFactory)initCtx.lookup("java:comp/env/jms/ConnectionFactory");                
            Connection connection = connectionFactory.createConnection();       
            connection.start();                
            javax.jms.Session session = connection.createSession(true, 0);                 
            // Look up for destination and set listener                
            Queue queue = (Queue) initCtx.lookup("java:comp/env/queue/hibernatesearch"); 
            MessageConsumer consumer = session.createConsumer(queue);
            consumer.setMessageListener(this);           
        } catch (Exception ex) {                
            // Looging
        }      
    } 
}

and the Spring bean configuration (we apply Spring3 to our application) for SessionFactory:

Code:

<bean id="clientSessionFactory"
      class="org.springframework.orm.hibernate3.annotation.AnnotationSessionFactoryBean">
      <property name="dataSource" ref="clientDataSource" />
      <property name="packagesToScan" value="tc.model" />
      <property name="hibernateProperties">
         <props>
            <prop key="hibernate.dialect">org.hibernate.dialect.MySQLDialect</prop>
            <prop key="hibernate.show_sql">false</prop>
            <prop key="hibernate.format_sql">false</prop>
            <prop key="hibernate.generate_statistics">true</prop>
            <!--<prop key="hibernate.hbm2ddl.auto">validate</prop> -->            

            <!-- JMS configuration -->
            <prop key="hibernate.search.default.sourceBase">d:/lucene/mastervolume/lucenedirs/mastercopy</prop>
            <prop key="hibernate.search.default.refresh">10</prop>
            <!-- Master node configuration -->
            <prop key="hibernate.search.default.indexBase">d:/lucene/master/lucenedirs</prop>
            <prop key="hibernate.search.default.directory_provider">filesystem-master</prop>
            
            <!-- Slave nodes configuration -->
            <!-- <prop key="hibernate.search.default.indexBase">d:/lucene/slave/lucenedirs</prop> -->
            <!-- <prop key="hibernate.search.default.directory_provider">filesystem-slave</prop> -->
            <!-- <prop key="hibernate.search.worker.backend">jms</prop> -->
            <!-- <prop key="hibernate.search.worker.jms.connection_factory">java:comp/env/jms/ConnectionFactory</prop> -->
            <!-- <prop key="hibernate.search.worker.jms.queue">java:comp/env/queue/hibernatesearch</prop> -->
            <!-- <prop key="hibernate.search.worker.execution">async</prop> -->
            <!-- <prop key="hibernate.search.worker.thread_pool.size">2</prop> -->
            <!-- <prop key="hibernate.search.worker.buffer_queue.max">10</prop> -->
         </props>
      </property>
   </bean>

My problem is when the HSearchListener call the getSession() method to process messages from slave nodes, it throws null exception due to null value of sessionFactory. It seems the bean can not be wired up for some reason.
It's really great for me to receive your helps to overcome this problem in particular to find some other ways to setup Hibernate Search on Amazon EC2 in general.

Actually, I also raised my problem in the article http://community.jboss.org/wiki/ClusteringUsingActiveMQAndTomcat as a comment. I wanna post it here too in the hope to receive more helps.

Thanks a lot, everybody.

sanne.grinovero · **Posted:** Sat Jul 23, 2011 8:23 am

Hello, welcome.

If it's a Spring error I can't help you (not an expert!), but maybe the SessionFactory won't start because of some other configuration error.
I assume you didn't find error messages in the log?

Also, are you using windows machines on EC2? I noticed the windows-style path in the configuration of the Lucene indexes, it seems a bit odd.

dungleonhart · **Joined:** Sat Jul 23, 2011 12:28 am **Posts:** 8

Hi,

Thanks a lot for your answer.
Actually, the configurations in my post are for my local Windows machine. I just want to do some testings before deploying to Linux server on Amazon (sorry for making confusion there).

With the SessionFactory problem, maybe you were right to say it was a kind of configuration error. Unfortunately, I have not figured it out and came up with a workaround: copy and modify the source codes of AbstartJMSHibernateSearchController class, replace the getSession() method with some configuration hard codes.

Here is my new classes:

Code:

/* Copy codes from class AbstartJMSHibernateSearchController of Emmanuel Bernard */
public class HSearchJMSHibernateSearchController implements MessageListener {
   private static final Logger log = LoggerFactory.make();

   /**
    * Process the Hibernate Search work queues received
    */
   @SuppressWarnings("unchecked")
   public void onMessage(Message message) {
      if ( !( message instanceof ObjectMessage ) ) {
         log.error( "Incorrect message type: {}", message.getClass() );
         return;
      }
      ObjectMessage objectMessage = (ObjectMessage) message;
      List<LuceneWork> queue;
      try {
         queue = (List<LuceneWork>) objectMessage.getObject();
      }
      catch (JMSException e) {
         log.error( "Unable to retrieve object from message: " + message.getClass(), e );
         return;
      }
      catch (ClassCastException e) {
         log.error( "Illegal object retrieved from message", e );
         return;
      }
      Runnable worker = getWorker( queue );
      worker.run();
   }

   private Runnable getWorker(List<LuceneWork> queue) {
      Runnable processor = null;

       /* Modified code: Directly get the SearchFactoryImplementor */
      SearchFactoryImplementor factory = HSearchContextHelper.getSearchFactory();
      processor = factory.getBackendQueueProcessorFactory().getProcessor( queue );
      
      return processor;
   }

   @PostConstruct
   public void initialize() {
      //init the source copy process
      //TODO actually this is probably wrong since this is now part of the DP
   }

   @PreDestroy
   public void shutdown() {
      //stop the source copy process
      //TODO actually this is probably wrong since this is now part of the DP
   }
}

Code:

public abstract class HSearchContextHelper {

   public static SearchFactoryImplementor getSearchFactory() {
      FullTextIndexEventListener listener = new FullTextIndexEventListener(Installation.SINGLE_INSTANCE);      
      Configuration cfg = new Configuration();
      Properties properties = new Properties();
      
      properties.setProperty("hibernate.dialect", "org.hibernate.dialect.MySQLDialect");
      properties.setProperty("hibernate.show_sql", "false");
      properties.setProperty("hibernate.format_sql", "false");
      properties.setProperty("hibernate.generate_statistics", "true");
      properties.setProperty("hibernate.search.default.sourceBase", "/mnt/lucenedirs/master");
      properties.setProperty("hibernate.search.default.refresh", "30");
      properties.setProperty("hibernate.search.default.indexBase", "/usr/lucenedirs/master");
      properties.setProperty("hibernate.search.default.directory_provider", "filesystem-master");
      cfg.setProperties(properties);
      
      cfg.addAnnotatedClass(Requirement.class);
      cfg.addAnnotatedClass(TestCase.class);
      cfg.addAnnotatedClass(Defect.class);   
      cfg.addAnnotatedClass(Project.class);
      
      cfg.buildMappings();
      cfg.buildSettings();

      listener.initialize(cfg);
      return listener.getSearchFactoryImplementor();
   }
}

Honestly, I'm feeling good with this workaround and not sure whether it makes any sense to you guys, but at least it seems to work fine. Do you have any advice for that?

Thanks a lot and Best regards,

sanne.grinovero · **Posted:** Fri Aug 05, 2011 1:49 pm

It seems you're initializing a new SearchFactory instance for every message you receive. That's going to be very slow, starting a SearchFactory might take close to a second, while reusing one could handle some thousand messages per second.. you're killing performance, but it might work.

I'd move that code in the initialize() method, and make sure the getWorker() reuses the same SearchFactory instance.