我正在探索Mule scatter-gather，以便在Mule fork和join 模式中并行处理流。我的Mule配置如下：
<flow name="fork" doc:name="fork">
<http:inbound-endpoint host="localhost" port="8090" path="mainPath" exchange-pattern="request-response" doc:name="HTTP"/>
<set-property propertyName="MULE_CORRELATION_GROUP_SIZE" value="2" doc:name="Property"/>
<!-- <all enableCorrelation="IF_NOT_SET" doc:name="All"> -->
<scatter-gather timeout="6000">
<processor-chain>
<async doc:name="Async">
<set-property propertyName="MULE_CORRELATION_SEQUENCE" value="1" doc:name="Property"/>
<flow-ref name="Flow1" doc:name="Flow Reference"/>
</async>
</processor-chain>
<async doc:name="Async">
<set-property propertyName="MULE_CORRELATION_SEQUENCE" value="2" doc:name="Property"/>
<flow-ref name="Flow2" doc:name="Flow Reference"/>
</async>
<!-- </all> -->
</scatter-gather>
<logger message="Main Flow" level="INFO" doc:name="Logger"/>
</flow>
<sub-flow name="Flow1" doc:name="Flow1">
<logger level="INFO" message="Flow1: processing started" doc:name="Logger"/>
<set-payload value="Flow1 Payload" doc:name="Set Payload"/>
<!-- Transformation payload -->
<logger level="INFO" message="Flow1: processing finished" doc:name="Logger"/>
<flow-ref name="Join-Flow" doc:name="Flow Reference"/>
</sub-flow>
<sub-flow name="Flow2" doc:name="Flow2">
<logger level="INFO" message="Flow2: processing started" doc:name="Logger"/>
<set-payload value="Flow2 Payload" doc:name="Set Payload"/>
<scripting:component doc:name="Groovy">
<scripting:script engine="Groovy"><![CDATA[sleep(2000); return message.payload;]]></scripting:script>
</scripting:component>
<!-- Transformation payload -->
<logger level="INFO" message="Flow2: processing finished" doc:name="Logger"/>
<flow-ref name="Join-Flow" doc:name="Flow Reference"/>
</sub-flow>
<sub-flow name="Join-Flow" doc:name="Join-Flow">
<collection-aggregator timeout="6000" failOnTimeout="true" doc:name="Collection Aggregator"/>
<combine-collections-transformer doc:name="Combine Collections"/>
<logger level="INFO" message="Combined Payload: #[message.payload]" doc:name="Logger"/>
<set-payload value="Soap XML Response" doc:name="Set Payload"/>
</sub-flow>
虽然我在记录器中获得聚合的有效负载，但分散和聚集正在抛出异常：
Exception stack is:
1. null (java.lang.UnsupportedOperationException)
org.mule.VoidMuleEvent:50 (null)
2. null (java.lang.UnsupportedOperationException). Message payload is of type: String (org.mule.api.MessagingException)
org.mule.execution.ExceptionToMessagingExceptionExecutionInterceptor:32 (http://www.mulesoft.org/docs/site/current3/apidocs/org/mule/api/MessagingException.html)
--------------------------------------------------------------------------------
Root Exception stack trace:
java.lang.UnsupportedOperationException
at org.mule.VoidMuleEvent.getMessage(VoidMuleEvent.java:50)
at org.mule.api.routing.AggregationContext$1.evaluate(AggregationContext.java:41)
at org.apache.commons.collections.CollectionUtils.select(CollectionUtils.java:517)
+ 3 more (set debug level logging or '-Dmule.verbose.exceptions=true' for everything)
********************************************************************************
INFO 2014-08-17 18:04:52,008 [[try1].fork.2.02] org.mule.api.processor.LoggerMessageProcessor: Flow2: processing started
INFO 2014-08-17 18:04:52,016 [[try1].fork.1.02] org.mule.api.processor.LoggerMessageProcessor: Flow1: processing started
INFO 2014-08-17 18:04:52,017 [[try1].fork.1.02] org.mule.api.processor.LoggerMessageProcessor: Flow1: processing finished
INFO 2014-08-17 18:04:54,309 [[try1].fork.2.02] org.mule.api.processor.LoggerMessageProcessor: Flow2: processing finished
INFO 2014-08-17 18:04:54,347 [[try1].fork.2.02] org.mule.api.processor.LoggerMessageProcessor: Combined Payload: [Flow1 Payload, Flow2 Payload]
一个有趣的事实是，如果我使用<all>路由器而不是分散 - 聚集，我就不会得到此异常。但我想<all>路由器按顺序处理流程而不是按Mule documentation并行处理流程。因此，我尝试在所有路由器上选择分散 - 聚集。但我不知道如何处理 scatter-gather 引发的上述异常。有什么出路吗？
更新流量： -
<flow name="fork" doc:name="fork">
<http:inbound-endpoint host="localhost" port="8090" path="mainPath" exchange-pattern="request-response" doc:name="HTTP"/>
<scatter-gather timeout="6000">
<processor-chain>
<flow-ref name="Flow1" doc:name="Flow Reference"/>
</processor-chain>
<flow-ref name="Flow2" doc:name="Flow Reference"/>
</scatter-gather>
<logger message="Main Flow" level="INFO" doc:name="Logger"/>
</flow>
<sub-flow name="Flow1" doc:name="Flow1">
<logger level="INFO" message="Flow1: processing started" doc:name="Logger"/>
<set-payload value="Flow1 Payload" doc:name="Set Payload"/>
<!-- Transformation payload -->
<logger level="INFO" message="Flow1: processing finished" doc:name="Logger"/>
<flow-ref name="Join-Flow" doc:name="Flow Reference"/>
</sub-flow>
<sub-flow name="Flow2" doc:name="Flow2">
<logger level="INFO" message="Flow2: processing started" doc:name="Logger"/>
<set-payload value="Flow2 Payload" doc:name="Set Payload"/>
<scripting:component doc:name="Groovy">
<scripting:script engine="Groovy"><![CDATA[sleep(2000); return message.payload;]]>
</scripting:script>
</scripting:component>
<!-- Transformation payload -->
<logger level="INFO" message="Flow2: processing finished" doc:name="Logger"/>
<flow-ref name="Join-Flow" doc:name="Flow Reference"/>
</sub-flow>
<sub-flow name="Join-Flow" doc:name="Join-Flow">
<collection-aggregator timeout="6000" failOnTimeout="true" doc:name="Collection Aggregator"/>
<combine-collections-transformer doc:name="Combine Collections"/>
<logger level="INFO" message="Combined Payload: #[message.payload]" doc:name="Logger"/>
<set-payload value="Soap XML Response" doc:name="Set Payload"/>
</sub-flow>
EXCEPTION： -
[try1].ScatterGatherWorkManager.02] org.mule.api.processor.LoggerMessageProcessor: Flow2: processing started
INFO 2014-08-18 13:34:24,361 [[try1].ScatterGatherWorkManager.01] org.mule.api.processor.LoggerMessageProcessor: Flow1: processing started
INFO 2014-08-18 13:34:24,364 [[try1].ScatterGatherWorkManager.01] org.mule.api.processor.LoggerMessageProcessor: Flow1: processing finished
WARN 2014-08-18 13:34:24,366 [[try1].ScatterGatherWorkManager.01] org.mule.routing.correlation.CollectionCorrelatorCallback: Correlation Group Size not set, but correlation aggregator is being used. Message is being forwarded as is
INFO 2014-08-18 13:34:24,401 [[try1].ScatterGatherWorkManager.01] org.mule.api.processor.LoggerMessageProcessor: Combined Payload: [Flow1 Payload]
INFO 2014-08-18 13:34:26,615 [[try1].ScatterGatherWorkManager.02] org.mule.api.processor.LoggerMessageProcessor: Flow2: processing finished
ERROR 2014-08-18 13:34:26,625 [[try1].connector.http.mule.default.receiver.03] org.mule.exception.DefaultMessagingExceptionStrategy:
********************************************************************************
Message : null (java.lang.NullPointerException). Message payload is of type: String
Code : MULE_ERROR--2
--------------------------------------------------------------------------------
Exception stack is:
1. null (java.lang.NullPointerException)
org.mule.api.routing.AggregationContext$1:41 (null)
2. null (java.lang.NullPointerException). Message payload is of type: String (org.mule.api.MessagingException)
org.mule.execution.ExceptionToMessagingExceptionExecutionInterceptor:32 (http://www.mulesoft.org/docs/site/current3/apidocs/org/mule/api/MessagingException.html)
--------------------------------------------------------------------------------
Root Exception stack trace:
java.lang.NullPointerException
at org.mule.api.routing.AggregationContext$1.evaluate(AggregationContext.java:41)
at org.apache.commons.collections.CollectionUtils.select(CollectionUtils.java:517)
at org.apache.commons.collections.CollectionUtils.select(CollectionUtils.java:498)
+ 3 more (set debug level logging or '-Dmule.verbose.exceptions=true' for everything)
********************************************************************************
现在您可以看到它在删除set-property后没有合并两个流的两个有效负载.....它只获得第一个流的有效负载： - 组合有效负载：[Flow1 Payload] ..
答案 0 :(得分：3)
删除async中消息处理器周围的scatter-gather范围，这确实为您进行并行化，因此您不必这样做。
编辑：同时删除处理set-property属性的MULE_CORRELATION消息处理器。 scatter-gather应该为你做到这一点。
EDIT2 ：您可以移除单processor-chain周围的flow-ref：它无用。
此外，似乎对scatter-gather的作用存在深刻的误解：你强行使其子流响应收敛到单个Join-Flow，在那里聚合东西，有效地绕过和重新实现scatter-gather建议为您做的所有事情。
将flow-ref移至Join-Flow，移除Join-Flow，然后将其处理逻辑（不是聚合器）放在scatter-gather之后。
答案 1 :(得分：1)
根据David的建议，最终的工作解决方案是： -
indices