Re: Updating a Field ( number (6) ) in 300.000 Rows cause 200 M RBS

This is wrong - although it might have been true for the BI file in version 5

The easy way to prove this (v8) is to look at v$transaction whilst doing a couple
of small update/deletes without committing.

There are two columns to watch:

     USED_UBLK
     USED_UREC

The number of rollback blocks used, and the number of rollback records written.

If you want to see what Oracle actually writes in the rollback, look at the

    START_UBAFIL
    START_UBABLK
from the same view, convert to a database block address and dump it (see my web site). The START_UBAREC tells you which undo record in the dump your transaction start on.

--

Jonathan Lewis
Yet another Oracle-related web site: www.jlcomp.demon.co.uk

Piotr Kolodziej wrote in message <7kcst2$ei1$1_at_korweta.task.gda.pl>...
>Karl Reitschuster wrote in message <37695F79.8017048C_at_t-online.de>...
>>What is with the Rest? or does Oracle put the whole row to the RBS?
>
>In rollbacks Oracle stores __whole__ undo blocks, not change vectors.
>So in the worst case rollback may contain all previous table and
>indexes blocks.
>HTH,
>--
>Piotr Kolodziej pkol_at_otago.gda.pl
>Just my private opinion.
>
Received on Fri Jun 18 1999 - 03:22:54 CDT