activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From arjun <>
Subject Re: Activemq producer locks
Date Sat, 05 Jan 2013 22:54:06 GMT
This is the background of the issue - 
Earlier the embedded broker in the tomcat used to be the default one that
gets created when you make a vm connection. So the memory limit used to be
64 mb and producer flow control used to be enabled. Also, there was only 1
camel consumer. With that state, once every 2 to 3 weeks, we used to have
huge log files all with the Producer flow control warning  message. This
happened for about 4 months or so. When we inspected the broker, we would
find that the queue had 65536 messages and the camel consumer would not
consume anything(because the memory limits were reached and producer flow
control had kicked in). At that point, we were not sure whether the consumer
stopped consuming or whether it the was the load(the load being so high that
consumer was not able to pull out messages as quickly as the producer
putting in the messages). So, we increased the number of consumers to 3,
increased the memory limit to 128 mb and disabled producer flow control. We
also added some jmx monitoring on the vm broker queue. With these settings,
just 2 days after we went into production, we saw that one consumer stopped
consuming messages(our jmx monitoring showed that there were 10 expired
messages and that's how we got alerted). We did a restart of the tomcat and
the next day again 1 consumer stopped consuming. After that this issue
hasn't occurred for about 2 - 3 weeks now.

A bit more info about our remote broker - we have three instances all
pointing to the same DB and we use failover transport. Also to note, the
remote broker had not failed over - while all the issues were happening 
with the tomcat, the same instance of the remote broker was active)

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

View raw message