You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@activemq.apache.org by jgogstad <jo...@gmail.com> on 2014/05/22 13:07:02 UTC
ActiveMQ producer hangs, consumers MessageCountAwaitingAcknowledge
increases
Hi,
I got an embedded broker with a producer in the same JVM using the
VM-transport. Out of the blue, this producer some times stop with no clues
as to why in the log. The symptoms are always the same though
* There one external consumer to a topic the producer is sending messages to
that got MessageCountAwaitingAcknowledge high and increasing. The consumer
has prefetch of Short.MAX_VALUE and MessageCountAwaitingAcknowledge is
typically around 10 000 - 20 000 messages. The MemoryUsageByteCount of the
topic is also steadily increasing.
Once I disconnect the consumer (by killing its transport, STOMP) my producer
starts working again.
1. Can anyone help me shed some lights on this?
2. As a sidenote, why is MemoryUsageByteCount increasing when a consumer
doesn't send acks and MessageCountAwaitingAcknowledge is less than the
prefetchSize of the consumer. There should be no reason to hold messages in
the broker in this situation?
--
Jostein
--
View this message in context: http://activemq.2283324.n4.nabble.com/ActiveMQ-producer-hangs-consumers-MessageCountAwaitingAcknowledge-increases-tp4681327.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
Re: ActiveMQ producer hangs, consumers
MessageCountAwaitingAcknowledge increases
Posted by jgogstad <jo...@gmail.com>.
In case anyone counters the same problem. Here's the actual problem and a
solution for my problem:
Problem:
Zombie threads holding open TCP connections to the broker on the STOMP
transport. The broker caches all messages going out to these consumers and
run out of memory. When memory usages hits 70-90% of the configured
memoryUsage in activemq.xml, all other producers are blocked with nothing
special logged.
Solution:
Use AbortSlowAckConsumerStrategy to kill off connections not sending ACKs.
See [1] and [2] for documentation.
Opened AMQ-5201 [3] to address missing logging and different behaviour in
STOMP and OpenWire transports.
[1]
http://planet.jboss.org/post/coming_in_activemq_5_9_a_new_way_to_abort_slow_consumers
[2]
https://access.redhat.com/site/documentation/en-US/Red_Hat_JBoss_A-MQ/6.0/html/Tuning_Guide/files/TuningSlowConsumers.html
[3] https://issues.apache.org/jira/browse/AMQ-5201
--
View this message in context: http://activemq.2283324.n4.nabble.com/ActiveMQ-producer-hangs-consumers-MessageCountAwaitingAcknowledge-increases-tp4681327p4681470.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.