Clustered Application setup with 3 managed nodes
EJB session replication enabled, passivation set to (24hrs or max of 200)
There was an out of memory in the managed server instance (with max heap 2.25GB), the heaps collected the time showed large 1.6 GB object of weblogic.ejb.container.swap.ReplicatedMemorySwap associated with one of the statefull session beans.
A further monitoring showed that Runtime MBeans are reporting that the Cached session statefull beans are being removed and passivated through the day.
A heap dump taken after 24hr passivation showed that the ReplicatedMemorySwap object had large secondaryBeanMap (ConcurrentHashMap). At the same time the RunTimeMBean reported only 5 session beans in the cache.
we also noticed that the size of actual beans were around the size less than 1MB, but the replicated swap was in the order of 300 MB
would be great to know if any one else had issues with replication on the statefull session beans in clusters
It seems to be the BUG and currently I see one BUG is in process of the fix. Bug 14521870 OOM DUE TO HEAP USAGE BY REPLICATEDMEMORYSWAP.
Please open a ticket with weblogic support to get more details on this BUG.