Re: same query, different parsing users, different plans ...
Date: Thu, 5 Jun 2008 07:10:51 -0700 (PDT)
> Not without seeing the plans and possibly 10053 traces for both.
> Also, you're running 10.2.0.2 and some optimizer-related issues were
> fixed in 10.2.0.3 so possibly you should consider patching?
I have 10053's for each scenario - the telling thing in them is that Oracle doesn't even consider the index on V2/T2 when it chooses the bad plan - search the entire file for it, and all you get is the part that lists its stats.
On the good plan, you can see it trying the join scenarios and costing a full scan of T2 and then an Index probe (which wins by a mile - cost of 3/4 instead of 20000 odd).
If you think you will have time to look I can post the full query and 10053's - its a lot of stuff to post to here though. Received on Thu Jun 05 2008 - 09:10:51 CDT