Karel PiwkoMatthias Wessendorf Looking at PagingManagerImpl and JournalImpl, the objects which got really huge are maps which contain in-flight transactions. Looks like that something started to fail transactions so HornetQ could not finish them. New and new transactions were started but they failed as well (most likely) and this map got really huge causing OOME (OutOfMemoryError).

As HornetQ is storing those transactions in file journal as well then when it tries to load journal after restart then new OOME can be thrown as it's putting all "unfinished" transactions to this map again during initialization.

By default there is 5 minute timeout for those transactions and then they're removed. But in this case it seems that memory got full before those transactions could be timeouted.

At this moment this seems to be a symptom but not the cause. Any idea why the transactions are failing? What are the first error messages in server.log?

Matthias Wessendorf hornetq-jms.xml just defines which queue/topics should be deployed together with the deployment. I cannot find configuration of the server where this deployment is deployed. Could you help? Is there described what the deployment is doing?