The reason for piling up of messages on the queue is due to single threaded
consumers connected through a non-pooled connection factory.
The first thing I would do is use jencks to set up a pooled connection
factory for ActiveMQ. An example of this is given below
<jee:jndi-lookup id="transactionManager"
jndi-name="javax.transaction.TransactionManager" />
<amqpool:xa-pool xmlns:amqpool="http://jencks.org/amqpool/2.0";
id="xaConnectionFactory"
brokerURL="Your Broker URL"
transactionManager="#transactionManager"
maxConnections="8" />
The next thing you need is to bump up your concurrent consumers. The queues
should be emptied as quickly as possible as shown below
<jms:consumer service="gta:JMSConsumerService"
endpoint="gtaRequestReader"
targetService="gta:connectorRequestRecipients"
destinationName="gta.request"
connectionFactory="#connectionFactory"
concurrentConsumers="8" />
This should take care of the issue.
Cheers,
Ashwin...
java programmer-2 wrote:
>
> Hello,
> We have problem with performance - we use currently 3 service
> assemblies, which communicate each other. One of them is core system,
> two other are connectors & XML transformers which returns common
> object model data, from different webservices.
> Problem is that between core system and two other components we use
> ActiveMQ queue, and after logging some times we found that messages
> remains most of the request lifecycyle in that queue waiting for
> consuming. We have no idea currently what to check to solve that
> problem, maybe you can get us any hint?
>
> We have defined queues for consumer & provider in component:
> <jms:consumer service="gta:JMSConsumerService"
> endpoint="gtaRequestReader"
> targetService="gta:connectorRequestRecipients"
> destinationName="gta.request"
> connectionFactory="#connectionFactory" />
>
> <jms:consumer service="gta:JMSConsumerService"
> endpoint="gtaResponseReader"
> targetService="gta:jms"
> targetEndpoint="gtaConnectorResponseWriter"
> destinationName="gta.response"
> connectionFactory="#connectionFactory" />
>
> <jms:provider service="gta:JMSProviderService"
> endpoint="gtaRequestWriter"
> destinationName="gta.request"
> connectionFactory="#connectionFactory" />
>
> <jms:provider service="gta:JMSProviderService"
> endpoint="gtaResponseWriter"
> destinationName="gta.response"
> connectionFactory="#connectionFactory" />
>
> <amq:connectionFactory id="connectionFactory"
> brokerURL="tcp://localhost:61616" />
>
> and routing in camel as:
> something().to("jbi:endpoint:http://www.test.com/JMSProviderService/gtaRequestWriter";);
> from("jbi:endpoint:http://www.test.com/jms/gtaConnectorResponseWriter";).tosomethingelse()
>
> Logged time is very huge between those two elements to and from, for
> the whole request time. Maybe there is need of some configuring how
> many elements should be consumed from queue in same time e.g. default
> is 1, so we have got sequencing order of consuming (one by one).
>
> Best regards,
> Adrian
>
>
--
View this message in context:
http://www.nabble.com/Poor-performance-over-queues--tp20579184p20581977.html
Sent from the ServiceMix - User mailing list archive at Nabble.com.
