Re: consistent reads and unique key updates
From: <jonathan_at_jlcomp.demon.co.uk>
Date: Wed, 25 Apr 2012 23:20:06 +0100
Message-Id: <E1SNAZC-0003NY-4X_at_pr-webmail-2.demon.net>
exriscer_at_gmail.com wrote:
> ok, just rerun the test but same output :-S
> end of the test and final query session statistics:
Date: Wed, 25 Apr 2012 23:20:06 +0100
Message-Id: <E1SNAZC-0003NY-4X_at_pr-webmail-2.demon.net>
exriscer_at_gmail.com wrote:
> ok, just rerun the test but same output :-S
> end of the test and final query session statistics:
I don't think I've seen a reference to the version number - but I may have missed some posts; I don't think you've shown any execution plans for the query or update.
However, one possibility is this: the update may have used the "switch current to new buffer mechanism" to clone the current copy, switching the original copy to state CR. If the first session then accessed the original it would be in the correct state for it to read without applying any undo.
In earlier versions of Oracle I think the update would have to do a tablescan for the 'switch current' to occur, and the first session would have had to do a "set transaction read only".
Regards
Jonathan Lewis
-- http://www.freelists.org/webpage/oracle-lReceived on Wed Apr 25 2012 - 17:20:06 CDT