HornetQ messages lost for slow consumer

[surve.swapnil]

I am using HornetQ running in standalone mode and the client in JBOSS AS 5.1.0

I am running test case where I am sending 10000 objectmessage with id and value. The consumer is slow consumer which persist the id,value in table and then thread sleeps for 1 second.The connection factory is XA configured.

During the consumption I am restarting the JBoss. The behaviour I am seeing is on first shutdown, no of messages on queue + no of messages consumed are same as no of messages sent. (I am checking the no of message based on no of id,value pair got persisted in database). On Next restart I see the count is lesser. On subsequent restarts the count is significantly lower. So no of messages persisted are pretty less than no of messages sent. I am loosing few thousand messages. This is unexpected behaviour.

One of the log statement says

2012-03-09 19:39:28,155 DEBUG [org.hornetq.ra.inflow.HornetQActivation] (WorkManager(2)-2) Using context {java.naming.factory.initial=org.jnp.interfaces.NamingContextFactory, java.naming.factory.url.pkgs=org.jboss.naming:org.jnp.interfaces:org.jboss.naming:org.jnp.interfaces} for org.hornetq.ra.inflow.HornetQActivationSpec(ra=org.hornetq.ra.HornetQResourceAdapter@6ab726 destination=queue/slowConsumerQueue destinationType=javax.jms.Queue ack=Auto-acknowledge durable=false clientID=null user=null maxSession=10)

But I have configured queue to be durable. (its default value)

What can be wrong here? I am attaching relevant files here.

[surve.swapnil]

additional details

HornetQ version is 2.5.0 final.

When I restart HornetQ in between (after bringing down the jboss), the no of messages on queue after broker restart are same as them before broker restart.

[clebert.suconic]

That durable on activation would be used only if this was a topic subscription.. which is not the case.

Can you try this on AS7? (just as a test)?

Make sure you enable persistence on the standalone.xml.

You can also take a look on what PrintData is showing. (If these messages are being ACKed or not):

java -Xmx2G -cp /path-to-jars/hornetq-logging.jar:/path-to-jars/hornetq-core.jar:/path-to-jars/netty.jar org.hornetq.core.persistence.impl.journal.PrintData /path-to-journal/bindins /path-to-journal/journal

This should tell you excatly what you have inside the journal.

Also: it was not clear to me if you're using the MDB in-vm or remotely.. it seems you removed the in-vm acceptor.

Take a look also at the consumer-window-size. It seems you need to disable client buffering.

[surve.swapnil]

Thanks for the reply.

Let me try out with setting consumer window size to 0. I was not aware of PringData. So couldnt find ACK logs. I will check on that.

I have remote MDB listening to queue. Since HornetQ is running as standalone and JBoss connecting to standalone HornetQ using JCA resource adapter. Is something wrong with the config?

[surve.swapnil]

I tried with consumer window size as 0. Still facing same problem.

Attaching PrintData logs when JBoss was brought down.

I set following in ra.xml

<config-property>

        <description>The consumer window size</description>

        <config-property-name>ConsumerWindowSize</config-property-name>

        <config-property-type>java.lang.Integer</config-property-type>

        <config-property-value>0</config-property-value>

      </config-property>

Also on hornetq-jms.xml following is set.

<connection-factory name="NettyConnectionFactory">

      <xa>true</xa>

      <connectors>

         <connector-ref connector-name="netty"/>

      </connectors>

      <entries>

         <entry name="/XAConnectionFactory"/>

      </entries>

   <consumer-window-size>0</consumer-window-size>

   </connection-factory>

This is very abrupt behaviour. Am i missing some configurations? I am attaching PrintData logs.

[clebert.suconic]

You didn't post your bindings.. you probably used a wrong directory.

it seems something weird on this:

### Message Counts ###

message count=16008

message reference count

queue id 81589,count=7346

queue id 7,count=8660

prepared message count=0

But I need to look at the bindings to what you have.

Maybe it would worth to try a checkout of the branch and buildling it?

[surve.swapnil]

Attaching PrintData logs with small set of 100 messages. Even with small set the problem is same. If I am reading the logs correctly, all the messages have been acknowledged. However I didint get them on MDB. There are absolutely no other consumers to this queue.