Re: DBA Humor (Twisted, sick sort of stuff...)

From: Hans Forbrich <fuzzy.graybeard_at_gmail.com>
Date: Wed, 27 Mar 2013 12:15:25 -0600
Message-ID: <5153373D.1000107_at_gmail.com>



On 26/03/2013 3:24 PM, Christopher.Taylor2_at_parallon.net wrote:
> Now, having sized these things fairly well, and not having these issues before, I started poking around (of course).
>
> We're under some pretty tight deadlines right now and some of our devs were being really aggressive with their testing.
>
> In the last 1.5 days we've generated over 500GB of archivelogs in this development instance.
Delete generates undo - delete [most] of a huge table, and you end up with no data but lots of undo, and as a result, lots of redo.

Oermanent tables to manipulate temporary data generates undo and redo, and often get emptied at the end of the transaction. Global temp tables will help with that, eliminating the direct redo although they still generate undo and associated redo, but without the 'clean out' transaction undo tracking. Ultimate size of temporary or working data is 'zero', but still has undo and redo.

Index rebuilds generate undo and redo while not changing the data size.

So yup, I can see how you can generate much more undo and redo than you hold in live data - especially in a test or dev environment that doesn't use flashback database to reset things.

/Hans

--
http://www.freelists.org/webpage/oracle-l
Received on Wed Mar 27 2013 - 19:15:25 CET

Original text of this message