Failover Transport Hot Backup Connections Not Being Closed By Connection Pool
When The Primary Connection Is Idle
-----------------------------------------------------------------------------------------------------------------
Key: AMQ-3118
URL: https://issues.apache.org/jira/browse/AMQ-3118
Project: ActiveMQ
Issue Type: Bug
Components: Connector
Affects Versions: 5.3.2
Environment: Centos 5.5
ActiveMQ 5.3.2 producer connecting to ActiveMQ 5.4.2 broker
Reporter: Timothy Myer
I am currently using the following URL in my consumer to connect to my AMQ
broker:
failover://(tcp://host1:61616,tcp://host2:61616)?randomize=false&maxReconnectAttempts=2&startupMaxReconnectAttempts=100&initialReconnectDelay=1000&backup=true&timeout=3000
I have specified the URL parameter backup=true to enable fast failover.
I am also using the following connection factory configuration from Spring:
<bean id="feeds-jms-connection-factory"
class="org.apache.activemq.pool.XaPooledConnectionFactory"
destroy-method="stop">
<property name="connectionFactory">
<bean class="org.apache.activemq.spring.ActiveMQConnectionFactory">
<property name="brokerURL"
value="failover://(tcp://host1:61616,tcp://host2:61616)?randomize=false&maxReconnectAttempts=2&startupMaxReconnectAttempts=100&initialReconnectDelay=1000&backup=true&timeout=3000"
/>
<property name="redeliveryPolicy">
<bean class="org.apache.activemq.RedeliveryPolicy">
<property name="maximumRedeliveries" value="3" />
<property name="initialRedeliveryDelay" value="100" />
</bean>
</property>
</bean>
</property>
</bean>
The problem is this: currently, my producers are not working at a very high
load, so my primary connections are in an idle state for more than 30 seconds,
which means that the connection pool is expiring them. Unfortunately, the pool
is not closing the hot backup connections (I can see this through either the
JMX console or by running netstat on the producer and broker boxes and seeing a
large number of open connections between the producer and broker). After a few
days, the broker is running out of file handles and the broker's CPU is pegged
because of all the persistent connections.
It seems that it should be the responsibility of the Pooled Connection Factory
to close any persistent backup connections (or perhaps it should be the
responsibility of whichever component creates the hot backup connection in the
first place). In any case, this seems to be a resource leak.
My only option right now is not to use hot backups and to let those backup
connections be created dynamically.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
