Customer Impact: Server can crash on OutOfMemoryError as maximum amount of memory allocated for each queue (defined by max-size-bytes attribute) is not respected. For administrator is not possible to estimate how much memory server can consume.

Test Scenario:

• Start EAP 7.1.0.DR19 with 2GB heap (-Xmx2g)
  • max-size-bytes is set to 1GB to all destinations
• Send 1 million of 1KB messages to queue InQueue (so ~1GB of messages)

Result: Server crashes on OOME during sending of messages.

Server crashed on OOME when ~659k messages were sent which should correspond to ~659 MB of messages in heap. Size of Artemis journal is 831 MB. Paging did not trigger so all messages are in "journal" directory. If server is started with larger heap (-Xmx3g) then server starts but consumes slightly more than 2 GB of heap after loading journal.

So every message in heap consumes more than double of it's size which is not the biggest issue. Worse thing is that max-size-bytes which is 1GB did not start paging when messages in memory for one queue exceeded 1GB of memory and let server crash on OutOfMemoryError.

There are 2 things to fix:

• trigger paging/block/fail/drop mode when memory allocated by one queue exceeded max-size-bytes
• why messages consumes so much of memory (more than double of their payload)

This is not regression.