I am trying to use spring integration to do a Gateway --> Splitter-->ServiceActivator --> Aggregator Pattern in an event driven fashion backed by JMS . I Expect the service activator to be multi-threaded and any of the end points can be executed on a cluster and not necessarily the originating server . I could get this working in a single JVM without using JMS ( Using SI Channels ) but I understand that SI Channels will not help me scale horizontally i.e multiple VMs .
Here's the configuration I have so far
<int:gateway id="transactionGateway" default-reply-channel="transaction-reply"
default-request-channel="transaction-request" default-reply-timeout="10000"
service-interface="com.test.abc.integration.service.ProcessGateway">
</int:gateway>
<int-jms:outbound-gateway id="transactionJMSGateway"
correlation-key="JMSCorrelationID" request-channel="transaction-request"
request-destination="transactionInputQueue" reply-channel="transaction-reply"
reply-destination="transactionOutputQueue" extract-reply-payload="true"
extract-request-payload="true">
<int-jms:reply-listener
max-concurrent-consumers="20" receive-timeout="5000"
max-messages-per-task="1" />
</int-jms:outbound-gateway>
<!-- Inbound Gateway for Splitter -->
<int-jms:inbound-gateway id="splitterGateWay"
request-destination="transactionInputQueue" request-channel="splitter-input"
reply-channel="splitter-output" concurrent-consumers="1"
default-reply-destination="processInputQueue"
max-concurrent-consumers="1" extract-reply-payload="true"
correlation-key="JMSCorrelationID" extract-request-payload="true" />
<!-- Inbound Gateway Invokes Service Activator and Sends response back to
the channel -->
<int-jms:inbound-gateway id="seriveActivatorGateway"
request-destination="processInputQueue" request-channel="process-input"
reply-channel="process-output" concurrent-consumers="1"
default-reply-destination="processOutputQueue"
max-concurrent-consumers="1" extract-reply-payload="true"
correlation-key="JMSCorrelationID" extract-request-payload="true"
max-messages-per-task="1" />
<int-jms:inbound-gateway id="aggregatorGateway"
request-destination="processOutputQueue" request-channel="aggregator-input"
reply-channel="aggregator-output" concurrent-consumers="1"
default-reply-destination="transactionOutputQueue"
max-concurrent-consumers="1" extract-reply-payload="true"
extract-request-payload="true" max-messages-per-task="1"
correlation-key="JMSCorrelationID" />
<int:splitter id="transactionSplitter" input-channel="splitter-input"
ref="processSplitter" output-channel="splitter-output">
</int:splitter>
<int:service-activator id="jbpmServiceActivator"
input-channel="process-input" ref="jbpmService" requires-reply="true"
output-channel="process-output">
</int:service-activator>
<int:aggregator id="transactionAggregator"
input-channel="aggregator-input" method="aggregate" ref="processAggregator"
output-channel="aggregator-output" message-store="processResultMessageStore"
send-partial-result-on-expiry="false">
</int:aggregator>
Before using gateway I tried using JMS backed Channels and that approach was't successful either . The problem I am facing now is that the Splitter now reply back to the transactionOutputQueue . I tried playing around with jms:header-enricher without much success . I feel that my approach to the problem /SI might have fundamental flaw . Any help /guidance is highly appreciated .
Also , in the code snippet I have provided above use a simple in memory aggregator , I understand that If I need to get this working across the cluster I might need a JDBC backed Aggregator but for the for now , I am trying to get this pattern working on a single VM
Here's the updated working configuration based on Gary's Comment
<bean id="processOutputQueue" class="org.apache.activemq.command.ActiveMQQueue">
<constructor-arg value="test.com.abc.process.output" />
</bean>
<bean id="transactionOutputQueue" class="org.apache.activemq.command.ActiveMQQueue">
<constructor-arg value="test.com.abc.transaction.result" />
</bean>
<bean id="transactionInputQueue" class="org.apache.activemq.command.ActiveMQQueue">
<constructor-arg value="test.com.abc.transaction.input" />
</bean>
<int:gateway id="transactionGateway"
default-request-channel="transaction-request" default-reply-timeout="10000"
default-reply-channel="aggregator-output"
service-interface="com.test.abc.integration.service.ProcessGateway">
</int:gateway>
<int:splitter id="transactionSplitter" input-channel="transaction-request"
ref="processSplitter" output-channel="splitter-output">
</int:splitter>
<int-jms:outbound-gateway id="splitterJMSGateway"
correlation-key="JMSCorrelationID" request-channel="splitter-output"
request-destination="processInputQueue" reply-channel="aggregator-input"
reply-destination="processOutputQueue" extract-request-payload="true"
extract-reply-payload="true">
<int-jms:reply-listener
max-concurrent-consumers="20" receive-timeout="5000" />
</int-jms:outbound-gateway>
<!-- Inbound Gateway Invokes Service Activator and Sends response back to
the channel -->
<int-jms:inbound-gateway id="seriveActivatorGateway"
request-destination="processInputQueue" request-channel="process-input"
reply-channel="process-output" default-reply-destination="processOutputQueue"
concurrent-consumers="5" max-concurrent-consumers="10"
extract-reply-payload="true" correlation-key="JMSCorrelationID"
extract-request-payload="true" max-messages-per-task="1" />
<int:service-activator id="jbpmServiceActivator"
input-channel="process-input" ref="jbpmService" requires-reply="true"
output-channel="process-output">
</int:service-activator>
<int:aggregator id="transactionAggregator"
input-channel="aggregator-input" ref="processAggregator"
output-channel="aggregator-output" message-store="processResultMessageStore"
send-partial-result-on-expiry="false">
</int:aggregator>
<bean id="processResultMessageStore"
class="org.springframework.integration.store.SimpleMessageStore" />
<bean id="processResultMessageStoreReaper"
class="org.springframework.integration.store.MessageGroupStoreReaper">
<property name="messageGroupStore" ref="processResultMessageStore" />
<property name="timeout" value="5000" />
</bean>
<task:scheduled-tasks>
<task:scheduled ref="processResultMessageStoreReaper"
method="run" fixed-rate="1000" />
</task:scheduled-tasks>
<int:logging-channel-adapter id="logger"
level="DEBUG" log-full-message="true" />
<int-stream:stdout-channel-adapter
id="stdoutAdapter" channel="logger" />
I limited the JMS pipeline only to the Service Activator , which is what I originally wanted .
The only question I have based on the above approach is that do I need to have my Aggregator backed by a database even if I use this across multiple VMS ( Since the JMS gateway in front of it make sure that it receives only the messages that have valid correlation ID ?)
Regards ,
