Recently I attempted, against I suppose my better instincts, to try an in-place upgrade on a small and simple Wss 3.0 farm consisting of 1 WFE server and 1 SQL server with about 400 MB of data. Having read the marketing literature and been assured that the upgrade story was much improved over the pervious version, i gave it a try. Now of course I first dutifully baked up the virtual servers and the site collections and the sql databases. Then I ran the in-place upgrade on the development box. The upgrade ran for about 5 minutes at most and completed completely successfully.
Then I ran it on the production server. It returned with 5 errors. The central admin site semi-upgraded with a broken master page. The rest of the portal completely failed and went down…404. The only thing in the logs was a mention of an API call. I had to restore the portal from backups. Is there any more tense moment in all of IT then when you get to find out how good your backups really are? Hyper-V server snapshotting, Thank You.
So I reran using the database attach method. The method I have always used, even back in 2007, and it has never let me down. Ran like a charm, upgraded everything and I am good to go.
The Doctor’s diagnosis. Don’t use the in place upgrade for production. I have the same opinion of it as Leo Getz does of the drive thru (Google it if you must :))