> The fix was to identify the process holding the latch and kill it. > MetaLink gives some information on how to identify a process holding a > latch (the v$ views do not, as it is never supposed to be held long > enough to do you any good). Unfortunately, I don't have the exact query > I used right at hand. Event 'hangahalyze' (Metalink Note:215858.1) can be used to identify any kind of locks (including latches) in database. Trace file for this event is human readable. Also I wrote a script ( http://oracledba.ru/hanganalyze/hanganalyze.py ) which generates trace file for 'hanganalyze' event, parses it and display a tree of locks in database. -- Egor http://www.oracledba.ru/orasrp/ Free Oracle Session Resource Profiler -- //www.freelists.org/webpage/oracle-l