On 3 Mar 2008, at 14:31, Rob Davies wrote:
On 29 Feb 2008, at 22:08, Wladimir Safonov wrote:
Hello,
as I found in the older mailings some people experienced the same
problem as I do in the moment. Seems like problem persisted already
in the earlier versions of the broker (< 4.0), but no solution
approach was found so far. So in my situation, I have a couple of
topics, some of which stay under relatively high message load (>10
messages pro sec) from the system start. After 1-2 hours a random
producer hangs up with the following stack trace:
sun.misc.Unsafe.park(Native Method)
java.util.concurrent.locks.LockSupport.park(LockSupport.java:118)
java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(AbstractQueuedSynchronizer.java:1841)
java.util.concurrent.ArrayBlockingQueue.take(ArrayBlockingQueue.java:341)
org.apache.activemq.transport.FutureResponse.getResult(FutureResponse.java:40)
org.apache.activemq.transport.ResponseCorrelator.request(ResponseCorrelator.java:76)
org.apache.activemq.ActiveMQConnection.syncSendPacket(ActiveMQConnection.java:1172)
org.apache.activemq.ActiveMQSession.send(ActiveMQSession.java:1640)
org.apache.activemq.ActiveMQMessageProducer.send(ActiveMQMessageProducer.java:226)
org.apache.activemq.ActiveMQMessageProducerSupport.send(ActiveMQMessageProducerSupport.java:240)
This can also happen to several producers at the same time. As I read
in the preceding posts, such behavior is possible if the consumer
fails. As I use failover, I can't say for sure if the consumer failed
just before the problem occurred. Anyway after the consumer has
recovered, the publisher should definitely receive some response,
which does not happen. Generally speaking, such broker strategy is
kind of odd, because possible listener crash & recover situations
should not influence the work of the producers - they should be
really decoupled.
I checked already this page:
http://activemq.apache.org/my-producer-blocks.html. But in my case
the problem also occurs on the empty topics which process a couple of
messages pro minute, whose consumer also fails now and then. So
current topic usage cannot be the reason for this behavior.
I'm currently using AMQ 5.0, but I had the same problem with AMQ 4.1.1.
Does anybody know how such broker behavior can be influenced or if
not, will there be any fixes regarding this bug in the future releases?
Appreciate any help!
Thanks a lot in advance!
Best regards,
Wladimir
Looks like your producer is waiting for resources (memory) to be freed
up on the broker. Could you try a 5.1-SNAPSHOT ?
Thanks for your advice! I've tried it with latest SNAPSHOT version, but
unfortunately producer still hangs up with the same stack trace after
1-2 hours of work. It doesn't occur to an overloaded topic, but rather
to a infrequently used one with a throughput 1 msg/min. What I also
noticed, from the two consumers on that topic only one crashes and
recovers, whereas the other one stays connected all the time. So merely
the recover process of a single consumer makes the broker block the
corresponding producer.
Do you have any other ideas how to fix this issue? Any workaround perhaps?
Thanks for any attempt to help in advance!
