Issue with ActiveMQXAConnectionFactory and Fuse Fabric

[cgeer_cxt]

I have an existing application running on the vanilla Apache ServiceMix stack that I'm trying to get running on a JBoss Fuse ESB release using Fabric and having an issue with A-MQ using XA Transactions. So my baseline is that this application works fine on my current platform. When moving to Fabric I changed my connection string from connecting to an AMQ broker in the same VM to using the Fabric discovery (i.e. discovery:(fabric:default)). Whenever I use an ActiveMQXAConnectionFactory and I try to send a message the application just hangs forever. If I change it to use an ActiveMQConnectionFactory it all works as a champ (but I don't have XA transaction support which isn't what I want). Here are my two bean configs:

    <bean id="activemqConnectionFactory" class="org.apache.activemq.ActiveMQConnectionFactory">

       <property name="brokerURL" value="discovery:(fabric:default)" />

        <property name="userName" value="mqconnection" />

        <property name="password" value="password" />

    </bean>

    <bean id="activemqConnectionFactoryXa" class="org.apache.activemq.ActiveMQXAConnectionFactory">

        <property name="brokerURL" value="discovery:(fabric:default)" />

        <property name="userName" value="mqconnection" />

        <property name="password" value="password" />

        <property name="redeliveryPolicy">

            <bean class="org.apache.activemq.RedeliveryPolicy">

                <property name="maximumRedeliveries" value="0"/>

            </bean>

        </property>

    </bean>

Something odd in the logs is that when I use the XA connection I start seeing messages like this which I don't see if I'm just using the non-XA connection:

2013-04-22 05:55:07,000 | DEBUG | ActiveMQ Task-1  | FailoverTransport                | sport.failover.FailoverTransport  786 | 2242 - org.apache.activemq.activemq-osgi - 5.8.0.redhat-60024 | urlList connectionList:[tcp://localhost:61616], from: [tcp://localhost:61616]

2013-04-22 05:55:07,001 | DEBUG | ActiveMQ Task-1  | FailoverTransport                | sport.failover.FailoverTransport  990 | 2242 - org.apache.activemq.activemq-osgi - 5.8.0.redhat-60024 | Attempting  35th  connect to: tcp://localhost:61616

2013-04-22 05:55:07,002 | DEBUG | ActiveMQ Task-1  | FailoverTransport                | sport.failover.FailoverTransport 1040 | 2242 - org.apache.activemq.activemq-osgi - 5.8.0.redhat-60024 | Connect fail to: tcp://localhost:61616, reason: java.net.ConnectException: Connection refused

2013-04-22 05:55:07,002 | DEBUG | ActiveMQ Task-1  | TcpTransport                     | ivemq.transport.tcp.TcpTransport  520 | 2242 - org.apache.activemq.activemq-osgi - 5.8.0.redhat-60024 | Stopping transport tcp://localhost:61616

I went through all my configs and code and can't find any references to localhost:61616 so I'm not sure where it's getting that from. My AMQ broker is being created by deploying the stock MQ profile to another container (so it's running in a different container than my application).

I think this is my last roadblock in getting some portion of my application running on Fabric so any help is appreciated.

Thanks,

Chris

[garytully]

that is odd.

can you increase logging such that we can see the application of the bean properties. It seems as if the <property name="brokerURL" value="discovery:(fabric:default)" /> is not being applied. But one would expect an error in that case. 

you are picking up the default broker url that is hard coded in the underlying factory: failover://tcp://localhost:61616

[cgeer_cxt]

Gary,

I finally figured out the issue but it doesn't really make sense to me based on the symptoms I was seeing. To fix the issue I changed my refernce to my XA Conenection Factory from

<reference id="jmsCF" interface="org.apache.activemq.pool.XaPooledConnectionFactory"/>

to

<reference xmlns:ext="http://aries.apache.org/blueprint/xmlns/blueprint-ext/v1.0.0" ext:proxy-method="classes" id="jmsCF" interface="org.apache.activemq.pool.XaPooledConnectionFactory"/>

This makes sence given I'm trying to reference a class and not an interface however there are two things that don't make sense.

1) Why I didn't get an error stating I couldn't reference a class (like I do in other places)

2) Why this would cause a random JMS connection to the default broker url which wasn't running

Chris

[garytully]

on 1) blueprint should error out. Can you raise an issue to track that so we can fix that for a future release.

on 2) I don't have an answer, something must be creating and referencing a connection factory, a little bit of remote debugging with a breakpoint in the creation method of ActiveMQConnectionFactory would track it down.