It seems your hprof dump already contains the information you need - you know which HashMap is unintentionally retaining the objects causing the leak. I would ask around in the WLS forums to see if someone can assist you in finding the root cause. If you want to reverse engineer what is responsible for allocating the objects being retained in the HashMap, you can fire up memleak and turn on allocation profiling for that particular type (weblogic.wsee.jaxws.ServerLateInitTube). That may however be a bit expensive. A Flight Recording may possibly help you, but only if the allocation of the leaked objects put into the hashmap is causing a considerable fraction of the total allocation pressure in the system.