骆驼分割和聚合失败，因为消息将发送到多个并发消费者

Question

我有一个简单的驼峰路由，它接受一个项目列表，将它们拆分，将每个元素发送到mq节点进行处理，然后通过聚合器将它们连接在一起。

非常接近合成消息处理器：http://camel.apache.org/composed-message-processor.html

但是我们注意到拆分后，camel会创建多个并发消费者？或者交换？因为消息被发送给多个消费者，他们永远不会完成。

列表：1,2,3,4

拆分：amq：：处理每个项目

骨料：

[Camel (camel-3) thread #41 - Aggregating 1 - Waiting on 3 more items
[Camel (camel-1) thread #16 - Aggregating 2 - Waiting on 3 more items
[Camel (camel-3) thread #49 - Aggregating 3 - Waiting on 2 more items
[Camel (camel-1) thread #15 - Aggregating 4 - Waiting on 2 more items

因此，camel生成了2个聚合器，每个聚合器都在等待4个项目，但每个聚合器只能获得两个。

骆驼路线：

<camelContext xmlns="http://camel.apache.org/schema/spring">

        <route> <!-- This route splits the reg request into it's items. Adding needed info to the message header.  -->
            <from uri="activemq:registration.splitByItemQueue" />  <!-- pick up the reg req -->
            <setHeader headerName="regReqId"> <!-- Need to store the Reg Req in the header  -->
                <simple>${body.registrationRequest.id}</simple>
            </setHeader>

            <split parallelProcessing="false" strategyRef="groupedExchangeAggregator"> <!-- Split the RegRequestInfo into it's individual requestItems (add, drop, etc) -->
                <method ref="requestSplitter"  method="split" />   <!-- does the actual splitting -->
                <setHeader headerName="JMSXGroupID"> <!-- This is CRITICAL. It is how we ensure valid seat check counts without db locking -->
                    <simple>FOID=${body.formatOfferingId}</simple>  <!-- grouping on the foid -->
                </setHeader>
                <to uri="activemq:registration.lprActionQueue"/> <!-- send to queue's for processing-->
            </split>
        </route>

        <route>    <!-- performs the registration + seat check -->
            <from uri="activemq:registration.lprActionQueue" />

            <bean ref="actionProcessor" method="process"/> <!-- go to the java code that makes all the decisions -->
            <to uri="activemq:registration.regReqItemJoinQueue"/> <!-- send to join queue's for final processing-->
        </route>

        <route>    <!-- This route joins items from the reg req item split. Once all items have completed, update state-->
            <from uri="activemq:registration.regReqItemJoinQueue" />  <!-- Every Reg Req Item will come here-->
            <aggregate strategyRef="groupedExchangeAggregator" ignoreInvalidCorrelationKeys="false" completionFromBatchConsumer="true"> <!-- take all the Reg Req Items an join them to their req -->
                <correlationExpression>
                    <header>regReqId</header> <!-- correlate on the regReqId we stored in the header -->
                </correlationExpression>

                <bean ref="actionProcessor" method="updateRegistrationRequestStatus"/> <!-- update status -->                   
            </aggregate>
        </route>
</camelContext>

<bean id="groupedExchangeAggregator" class="org.apache.camel.processor.aggregate.GroupedExchangeAggregationStrategy" />

在我的本地机器上，上述工作正常，但是当我们部署到我们的测试服务器时，一半的消息转到一个骆驼聚合器，一半转到另一个。导致没有一个完成。请注意，在下面的配置中，我们已将骆驼的并发消费者设置为1。

这是camel/activemq配置

<amq:broker useJmx="false" persistent="false">
        <amq:plugins>
            <amq:statisticsBrokerPlugin />
        </amq:plugins>
        <amq:transportConnectors>
            <amq:transportConnector uri="tcp://localhost:0" />
        </amq:transportConnectors>
    </amq:broker>

    <!-- Basic AMQ connection factory -->
    <amq:connectionFactory id="amqConnectionFactory" brokerURL="vm://localhost" />

    <!-- Wraps the AMQ connection factory in Spring's caching (ie: pooled) factory
         From the AMQ "Spring Support"-page: "You can use the PooledConnectionFactory for efficient pooling... or you
         can use the Spring JMS CachingConnectionFactory to achieve the same effect."
         See "Consuming JMS from inside Spring" at http://activemq.apache.org/spring-support.html
         Also see http://codedependents.com/2010/07/14/connectionfactories-and-caching-with-spring-and-activemq/

         Note: there are pros/cons to using Spring's caching factory vs Apache's PooledConnectionFactory; but, until
         we have more explicit reasons to favor one over the other, Spring's is less tightly-coupled to a specific
         AMQP-implementation.
         See http://stackoverflow.com/a/19594974
    -->
    <bean id="connectionFactory" class="org.springframework.jms.connection.CachingConnectionFactory">
        <constructor-arg ref="amqConnectionFactory"/>
        <property name="sessionCacheSize" value="1"/>
    </bean>

    <bean id="jmsTemplate" class="org.springframework.jms.core.JmsTemplate">
        <constructor-arg ref="connectionFactory" />
    </bean>

    <bean id="jmsConfig"
          class="org.apache.camel.component.jms.JmsConfiguration">
        <property name="connectionFactory" ref="connectionFactory"/>
        <property name="concurrentConsumers" value="1"/>
    </bean>

    <bean id="activemq"
          class="org.apache.activemq.camel.component.ActiveMQComponent">
        <property name="configuration" ref="jmsConfig"/>
    </bean>

Answer 1

原来我们有另一个Spring上下文/servlet导入我们的配置。我们相信这是问题所在。