what version were you using?
Can you try the latest 2.1.0.CR1?
Currently using version 2.0.0.GA
I've tried building 2.1.0.CR1 from the trunk, but no luck. There was no netty-3.2.0.CR1 in the maven repository at that time. I can see it is there now, so I will try to build it again.
At this very moment, Trunk == 2.1.0.CR1
I am continuing to get negative message counts using 2.1.1.Final
Any ideas on a cause/fix for this?
I don't know of any bug reports about this.
We would need something replicating your issue... (http://community.jboss.org/docs/DOC-14508)
Right now the only thing I could think is http://hornetq.sourceforge.net/docs/hornetq-2.1.0.Final/user-manual/en/html/undelivered-messages.html#configuring.delivery.count.persistence
Associated with some failure at your system. But I don't know anything about your system. As I said we would need some way to replicate your issue (a testcase for instance).
I seem to have the scenario which causes this.
When we move messages from one queue to another (programatically using the QueueControl over JMX), and the original queue has client side caching enabled. While not every time, this seems to cause the negative count, when the final cached messages are consumed. I've also noticed that the records that are read out of the original queue are sometimes duplicates of the messages that were put in the other queue via the moveMessages call, and some of the messages are not duplicates. Possibly related to the negative count...
One final note, I've discovered that when the message count is negative, we lose all messages until the count is >= 0
I had 4 subscribers receiving the same messages, 3 of which had different negative counts, no messages were received by the subscriber until the count got higher.
Sub1 = -2
Sub2 = 0
Sub3 = -8
Sub4 = -19
When I ran 10 messages, the subscribers processed the following number of messages:
Sub1 = 8
Sub2 = 10
Sub3 = 2
Sub4 = 0
This is a pretty big issue for us.
As I said earlier, do you have a testcase replicating this?
Can you create one?