This is normally due to your application logic not being fast enough.

What happens is that RFA has a background thread which reads messages from the network. It puts these messages into an internal queue where they sit until your application processes them (i.e your application calls the dispatch() method on your event queue). This internal queue is by default unbounded.

In other words: you are simply not calling dispatch() often enough and events are piling up. This will continue until you get an OutOfMemory error. Increasing the heap will be treating the symptom, not the cause.

Some suggestions:

• The dispatch() method returns a value. Use it! If you've made a call to dispatch() and it returns that there are still say 10000 messages on the queue, then you know that you are really falling behind. In a well-behaved application there should never be more than say 10-50 messages in that queue, IMO, and that's only when there's a burst. Otherwise you should be dispatching fast enough to keep the queue at almost zero length.
• Look for deadlocks in your implementation of com.reuters.rfa.common.Client.
• Look for slowness in your implementation of com.reuters.rfa.common.Client. Re-consider what you are doing in that loop? Something with a database, for example? Normally this will not fly as there will be too many messages for any database to keep up.
• Use VisualVM to profile your application, in particular the processing logic in your com.reuters.rfa.common.Client class.
• If your dispatching thread is stuck then use VisualVM to inspect where it is stuck.

Hope you can use the above.

Lars

Addicticks