Thanks, Tanel! Your addional explanation cleared it up for me again! > The "composite granule" > code isn't probably able to reclaim that memory back from buffer cache use > to shared pool - or doesn't realize that it would be good to shrink the > buffer cache... This possibly is the culprit of what I observed and called a memory leak. My expectation would be that at least when you get ORA-04031 errors, the instance should know that it would be better to reclaim space used by the buffer cache. Cheers and happy OOW! Marcus -- //www.freelists.org/webpage/oracle-l