> Plan on identifying issues in your test upgrades in your test > environment, opening iTARs as necessary and waiting for > fixes. We've hit 4 different errors upgrading from 9.2. to > 10.1 - some had workarounds, some didn't. Your schedule > should be appropriately unaggressive. > > I'm not saying that 10.1.0.4 is buggy. > I'm saying that upgrading from 9.2.0.5.x or 9.2.0.6 to > 10.1.0.4 is less than perfect. > (lets just say that its not a state function - its path > dependent and there be dragons along that path). > > exp/imp into a clean db if you can. > Hi Paul, as you mentioned 10g is not bugfree. I met some issues too: ORA-600 when importing from 8.1.7.0 (during index creation), crash of the migration assistant from 9.2.0.4 to 10.1.0.4 (minor issue since I was able to use the sql scripts), listener in 10.1.0.4 sometimes freezes due to a problem in ONS registration (this is critical but a workaround does exist). Plus some minor issues. My new tests on 10gR2 are not so good. Several issues with datapump. ORA-600 for unknow reasons (TARs still open) and problem with buffer locks. So far my personal list of most stable release (I never worked with anything older than 8.1.7.0): 10.1.0.4 (and 10.1.0.3), particularly on linux. 8.1.7.4 on almost every platform 9.2.0.4 (ok, I added this one only because I needed a third entry and I'll avoid this version on solaris). Most hated one: 9.0.x (I still have one as a metadata repository of a IAS). Special mention of my black list: 9.2.0.6 with RAC (first time I had to roleback a patchset!). My reason for 10g migration is not due to new features but only to stability (or better: instability of previous releases). Fabrizio -- //www.freelists.org/webpage/oracle-l