We also experience stuck messages in activemq 5.1, slow consumers an
jdbc persistance.
Searching for the reason and/or solution to this problem i stumbled last
week on this blog entry:
http://kovyrin.net/2009/01/23/activemq-tips-flow-control-and-stalled-producers-problem/
So we set up our testing environmen with activemq 5.2 (which should fix
several other problams like the annoying broken stats in
jconsole/webconsole), turned off flow control and redurced the queu
prefetch size to 0.
Well up to now no message got stuck.
Maybe it's also usefull for your environnment.
If it helps I'd be thankful for a short response on this
us...@activemq.apache.org.
Greetings,
Norbert
bwtaylor schrieb:
We experience this problem as well. It's a very serious issue, and I hope it
gets some attention. We don't have test code that can reproduce it, as it
seems intermittent. We can reproduce it when we get lucky by jamming a lot
of messages into queues with slow consumers. We've seen it with single or
multiple consumers. Typically we're using camel to consume the messages.
When we've seen this occur, using jconsole shows that the inflight count
remains low, indicating the stuck messages are not being dispatched. The
messages are in the database, so like others, bouncing the broker restores
the problem. We did manage to capture a heap dump during the failure. We
tried to inspect this, but didn't get much out of it. We're currently on
activemq 5.1. We started to try the 5.3 snapshot to see if we could
reproduce it with 5.3, but because of the oracle master/slave breakage I
just reported in the dev mailing list, that version doesn't work for us at
the moment.
