camel-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mark de Leon (JIRA)" <j...@apache.org>
Subject [jira] [Created] (CAMEL-11480) camel-rabbitmq - autorecovery creates additional channels
Date Thu, 29 Jun 2017 10:45:00 GMT
Mark de Leon created CAMEL-11480:
------------------------------------

             Summary: camel-rabbitmq - autorecovery creates additional channels
                 Key: CAMEL-11480
                 URL: https://issues.apache.org/jira/browse/CAMEL-11480
             Project: Camel
          Issue Type: Bug
          Components: camel-rabbitmq
    Affects Versions: 2.19.1
         Environment: rabbitmq server 3.6.2 running on cluster+HA mode (2 nodes), Spring client
application using DSL camel configuration
            Reporter: Mark de Leon
            Priority: Minor


While testing the automatic recovery for rabbitmq for network failure scenario, I noticed
that after recovery, the consumer/channel count is always increasing.

Consider the following codes:
<camel:route id="in_route">
			<camel:from uri="rabbitmq://rabbitmq/myexchange?connectionFactory=#rabbitMqCustomConnectionFactory&amp;addresses={{rabbitmq.nodes}}&amp;queue=myqueue"
/>
			<camel:to uri="log:message received?showBody=true" />
		</camel:route>

// Custom connection factory
<bean id="rabbitMqCustomConnectionFactory" class="com.rabbitmq.client.ConnectionFactory">
	<property name="username" value="${rabbitmq.username}" />
	<property name="password" value="${rabbitmq.password}" />
</bean>

When the application runs, I see 2 threads running:
* RabbitMQConsumer [camel-rabbitmq library]
* AMQConnection (RecoveryAwareAMQConnection) [amqp-client]

Killing the active rabbitmq server node triggers the auto recovery in the client.
* RabbitConsumer.handleShutdownSignal() successfully reconnects to the other rabbitmq node.

* At the same time, AutorecoveringConnection.beginAutomaticRecovery is triggered but with
some networkRecoveryInterval delay - recovers successfully any existing channels.

As a result, 2 channels/consumers are created - both are active. When there should only be
1 channel to be recovered.

Would it be possible to add a priority whether which recovery mechanism to use? Was thinking
of having the AMQConnection recovery execute first before the RabbitConsumer recovery since
the latter checks if an existing channel is open before creating a new connection.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message