We experienced a memory leak in qpidd that resulted in the daemon crashing. Memory utilization grew over a roughly 24-hour period before it finally reached the 32-bit 2G limit. Utilization is fairly light, it only does a few hundred messages a minute. Messages are very small, under 100 bytes (payload + routing key) each. All messages are done on topic queues on 3 exchanges. No XML or anything fancy, just a very simple pub-sub with no message persistence.
Error messages in the logs show std::bad_alloc being thrown for allocations of sizes 320 (shows up once at the beginning of the error messages) and 501 (shows up several times thereafter until the crash). Unfortunately, I don't have any other data to give, but here's a graph showing the allocation behavior:
--------------------------------------------------------------------- Apache Qpid - AMQP Messaging Implementation Project: http://qpid.apache.org Use/Interact: mailto:[email protected]
