I have recently been dealing with qpid (the last release M2 version) running out of memory. I'm running it on a linux box with 4 gigs of ram (2 gigs allocated to qpid).
I start sending data to qpid around 9:30 through the C# api and qpid starts with about 10 megs of memory use. Today, around 3:30, it used up its two gigs. As mentioned in a previous email, the only reason for this I can think of is that the queues are not getting dequeued fast enough (the clients are too slow) and they start to fill up. I spent some time going through the JMX monitor (VisualJVM), but, frankly, I can't tell which queues are the ones taking up all the memory. I have an hprof file (1.5 gig) as well which is generated when the jvm dies due to memory problems. I can't quite tell what to look for in the file either. Finally, I don’t think my clients are too slow. The largest amount of data I publish is less than 150 megs! How can I resolve this? If C# clients were keeping up with the trunk version of qpid, I'd just upgrade and see if the issue gets resolved :)
