I'd suggest that you disable producer flow control - see
http://activemq.apache.org/producer-flow-control.html
This will allow messages for a slow consumer to be off-lined to disk
cheers,
Rob
On Jan 5, 2008, at 12:19 AM, EJLeVin1 wrote:
We have been using 5.0 -- but today I tried seeing if the 5.1
SNAPSHOT would
be the fix, but that didn't make any difference. One thing we were
able to
solve -- we got the producer to continue sending messages even if
one of the
consuming applications was "paused". The way we did that was
through using
a SimpleDispatchPolicy; HOWEVER, this caused an out-of-memory issue
eventually because it seems that although the messages are getting
sent to
the other consumers that are not paused, they are seemingly also
getting
queued for the paused consumer. I am still not quite sure why this
would be
(as we are setting the pendingMessageLimitStrategy to 1? Any
thoughts?
Thanks for taking the time to help!
Hiram Chirino wrote:
What version you running with?
On Jan 4, 2008 1:22 AM, EJLeVin1 <[EMAIL PROTECTED]> wrote:
Hi,
First off I want to say I am sorry right up front if this is a
very
basic question, but I as well as others have been searching for a
solution
to this to no avail. We have run into the following situation in
testing
one of our new developments:
-Non-persistent TOPIC
-Very fast STOMP producer (>3000 msgs / sec)
-Multiple STOMP consumers
The problem we are running into is that this whole setup has been
working
excellent up to this point; however, we discovered a major issue
with
Slow
Consumers being handled. I know that there are many questions/
postings
regarding handling slow consumers, but yet we still are running
into this
issue, and I am beginning to wonder if it is currently how AMQ is
just
functionally built. Essentially what happens is if one of these
consumers
is slow (we throttled the receive through TCPMon) or a breakpoint
in hit
on
the consuming application, the producer will continue to send for
a time,
but eventually (and quickly) goes into a wait for the send-queue
of the
slow
consumer to have more space and no messages are sent to any of the
clients.
If the slow consumer disconnects or the breakpoint is done being
debugged,
everything goes back to normal, but not until that point. Below
is the
config file section I tried to put into place given the other forum
postings
/ AMQ articles I read; however, the same result occurred.
What I really just want to have happen is messages get delivered
to the
consumers on my topic as they come in, if a consumer is slow to take
messages and multiple messages come in while the slow delivery is
occurring,
the subsequent messages get delivered to the other consumers and
then
just
thrown away. I need to find a way to configure things where even if
consumers on my topic are not able to keep up, they will get
messages as
fast as they can consume them, but they will not interfere with
other
consumers getting the produced messages as well. We have also tried
different subscription params such as:
activemq.dispatchAsync
activemq.maximumPendingMessageLimit
By the way, when I am sending the messages to AMQ, I am setting the
message
to non-persistent, but that also had no effect.
<policyEntry topic="MyTopic.>">
<subscriptionRecoveryPolicy>
<noSubscriptionRecoveryPolicy/>
</subscriptionRecoveryPolicy>
<!-- lets force old messages to be discarded for slow
consumers
-->
<pendingMessageLimitStrategy>
<constantPendingMessageLimitStrategy limit="1"/>
</pendingMessageLimitStrategy>
</policyEntry>
</policyEntries>
Thanks in advance for your help, I really appreciate it.
-Eric
--
View this message in context:
http://www.nabble.com/%27Hopefully%27-Configuration-Issue-tp14611574s2354p14611574.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
--
Regards,
Hiram
Blog: http://hiramchirino.com
Open Source SOA
http://open.iona.com
--
View this message in context:
http://www.nabble.com/%27Hopefully%27-Configuration-Issue-tp14611574s2354p14627714.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
