2 things: 1. what patchset are you on? (gratuitous plug for Alert #68 patchsets, e.g. 9.2.0.5.3) 2. data loads - you may find that a lrage number of undo segments are brought online - and not released in the amount of time that you would expect, even with undo_retention set to 300 seconds. Check for posts by Dan Fink and Kirti Deshpande covering such issues. I believe that they both have papers up on the RMOUG site. hth. Paul On Fri, 17 Sep 2004 09:16:15 +0100, Paul Vincent <paul.vincent@xxxxxxxxx> wrote: > Now that we've finally got some 9i databases, I'm quite keen to switch = > from Manual to Automatic Undo Management. It all looks pretty = > straightforward, but I do have a couple of questions that the "How To" = > guides don't really address... > > 1. If I don't have any requirement for long-term retention of Undo data = > after transactions commit, so that the default retention of 900 seconds = > is quite sufficient, then what multiplier should I use when calculating = > the size of my Undo Tablespace relative to the size of the old RBS = > tablespace? Say I currently have 500Mb RBS tablespace, then should a = > 500Mb undo tablespace be adequate, or should I allow double (1Gb) or = > more? > > 2. What monitoring practices are advisable to check the adequacy of the = > undo tablespace provision? Can anyone point me at any webpages with = > articles about monitoring AUM activity, for instance? > > 3. Once I've switched from Manual to Auto, does the RBS tablespace then = > become completely superfluous? Can it be dropped? And what of the = > "rollback_segments" initialisation parameter? Do I remove this from the = > parameter file, or is it best left there "just in case"? > > Before I go diving in, in cavalier manner, I always like to check for = > "gotchas"! > > Regards, > > Paul Vincent > DBA > University of Central England -- //www.freelists.org/webpage/oracle-l