activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From dlaube <>
Subject Cannot consume messages on queue after master/slave failover in ActiveMQ 5.8.0.
Date Sat, 12 Oct 2013 01:17:20 GMT
I am encountering some issues with master/slave failover with ActiveMQ 5.8.0
running on RHEL and would appreciate some recommendations/help from the
community if possible. I have tried searching but nothing seems to resolve
my issue. Using the ActiveMQ example apps (per I setup the consumer and producer
using the failover transport and then walked through a failure on master.
Steps to reproduce, along with the commands are included below.

The consumer starts eating the messages that the producer sends to my
TEST.FOO queue. Next, when I kill the ActiveMQ master broker, I see that the
the producer reconnects to the slave/new-master and will continue sending
messages where it left off. However, the consumer re-connects but then fails
to receive any messages from the slave/new-master after failover. I see that
the TEST.FOO queue was (re)created on the slave/new-master after the
failover takes place. The consumer reports the following;  [java] 00:00:35
WARN  Transport (tcp://master_IP:61616) failed, reason: Connection reset, attempting to automatically
reconnect    [java] 00:00:39 INFO  Successfully reconnected to

Steps to reproduce:
1. Start ActiveMQ broker master  and slave
2. Start ActiveMQ producer and consumer ( parenthesis and ampersands are
escaped for good measure, otherwise BASH complains as expected )
ant producer
-Dtopic=false -DtimeToLive=30000 -Ddurable=false -Dmax=80000
ant consumer
-Dtopic=false -Ddurable=false -Dmax=80000
3. Kill ActiveMQ master and observe ActiveMQ slave takeover. Look for string
"Becoming the master on dataSource:
org.apache.commons.dbcp.BasicDataSource@2482c23d" in ActiveMQ logs located
on the slave.
4. Observe timeout/reconnect operations in both the consumer and producer
that was started in step #2. The producer will reconnect 

Has anyone encountered this issue before with a two-node master/slave
implementation of ActiveMQ? If so, what was the resolution?

Best regards,
-David Laube

View this message in context:
Sent from the ActiveMQ - User mailing list archive at

View raw message