Tracer start failed when starting multiple MqProducer in one process

[kangjiabang]

i want to use trace feature in 4.4.0.
when i init two defaultMQProducer ， use new DefaultMQProducer(producerId,true);
then i start to send message to test trace， trace not work well。

i find when producer start with trace on，it will produce new AsyncTraceDispatcher，in AsyncTraceDispatcher, it will invoke getAndCreateTraceProducer method.

private DefaultMQProducer getAndCreateTraceProducer(RPCHook rpcHook) { DefaultMQProducer traceProducerInstance = this.traceProducer; if (traceProducerInstance == null) { traceProducerInstance = new DefaultMQProducer(rpcHook); traceProducerInstance.setProducerGroup(TraceConstants.GROUP_NAME); traceProducerInstance.setSendMsgTimeout(5000); traceProducerInstance.setVipChannelEnabled(false); // The max size of message is 128K traceProducerInstance.setMaxMessageSize(maxMsgSize - 10 * 1000); } return traceProducerInstance; }
but the ProducerGroup is same for all trace Producer . when i start two MQProducer ,i will report error like following .
throw new MQClientException("The producer group[" + this.defaultMQProducer.getProducerGroup() + "] has been created before, specify another name please." + FAQUrl.suggestTodo(FAQUrl.GROUP_NAME_DUPLICATE_URL), null);

[ChaosYjh]

I think it is a bug too, but it seems that no one cares about this problem.@duhenglucky ,can you look at this question? Thanks~~

[duhenglucky]

@ChaosYjh try to setInstanceName when creating producer，and recommend to create one producer to send different topic messages, and it would be nice if you can proposal a PR to polish this issue.

[zongtanghu]

@ChaosYjh @kangjiabang We will check this issue in our local enviroment.

[ChaosYjh]

@ChaosYjh try to setInstanceName when creating producer，and recommend to create one producer to send different topic messages, and it would be nice if you can proposal a PR to polish this issue.
Set instanceName can resolve this problem,but it will create double tcp connection at least.
In our users, there are many application system use producer and consumer both ,and there are three or four consumerGroup,may be more.So I hope they use one clientInstance in common to create tcp connection as little as possible and save resource.
In AsyncTraceDispatcher,traceProducer specified producerGroup by TraceConstants.GROUP_NAME always, so I think can we specify a different producerGroup for each traceProducer and use one instanceName in common?

[ahuazhu]

@ChaosYjh Thanks, Set different ProducerGroups for consumer/producer will resolve the problem. I will create a PR to fix the bug.
#1275

[lnfdxwl]

@ahuazhu hi，when make release ，thanks

[duhenglucky]

fixed in the 4.6.1