activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rob Davies <rajdav...@gmail.com>
Subject Re: Large PendingQueueSize values for subscriptions on a queue with one message
Date Thu, 23 Aug 2007 10:02:35 GMT
could you build a junit test case to demonstrate this ? This would be  
most hopeful


cheers,

Rob

'Go Get Integrated -  ride the Camel! - http://activemq.apache.org/ 
camel/'
http://rajdavies.blogspot.com/



On Aug 23, 2007, at 9:22 AM, David Sitsky wrote:

> I have an application which is using the latest snapshot of  
> ActiveMQ 5.0.
>
> I have a master JVM process which sends an item to a work-item  
> queue.  I have worker JVM processes which pick up these items,  
> process them, and possibly create new items into the work-item  
> queue as a part of processing these messages.
>
> The master sends a new work-item once the workers have completed  
> all their work.
>
> I have a situation where if there is only one worker JVM process  
> (one consumer), everything works fine.  If there are two workers,  
> they complete all of their work, the master sends the next work  
> item to the queue, but they never get notified of this new message.
>
> I know the message is sent, via the JMX console, and that the  
> worker subscriptions are still active.
>
> What is strange is the worker's subscriptions have large  
> PendingQueueSize values, which seems odd, since they aren't  
> processing any more messages.  The queue size is 1 after the master  
> sends its new message.
>
> If I create a new worker process, it immediately gets the message  
> from the master!
>
> This happens with or without transactions, and with all sorts of  
> pre-fetch sizes, using the same or separate Connections.
>
> I've started to run the broker in my IDEA debugger to try and  
> understand what is going on - but the fact that the  
> PendingQueueSize are quite large, does this explain why the old  
> workers don't get new messages? Does this sound like a bug?
>
> Any suggestions on what I should look for?  I'm happy to dive into  
> the code, but would appreciate some guidance.
>
> Cheers,
> David


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message