Re: Amount of arc log differs in two identical db's ?

From: joel garry <joel-garry_at_home.com>
Date: Mon, 28 Dec 2009 10:23:08 -0800 (PST)
Message-ID: <0906ffa5-3feb-4fb8-96c9-5ab123c102d2_at_a21g2000yqc.googlegroups.com>



On Dec 28, 12:44 am, eh <timo.talja.em..._at_gmail.com> wrote:
> Hi,
> There are two identical oracle 9.2 db's running identical software and
> basically amount of transactions is same - still there's a huge
> difference between archive log amount generated in these db's.
> Logs generated from other db is something like four times more ....
> Not going more deeper in this case - what could result this kind of
> behaviour ???
>
> Thanks,
> BR,
> timo

Just throwing out some additional thoughts, as vsevolod has stated the obvious two:

How identical is identical? We need to know every exact difference, as sometimes a seemingly small difference (or the combo of a few) can make a big effect. Please show SGA, and show all init.ora parameters that are not default. Show us the physical memory, exact operating system version, hardware configuration.

Amount of transactions, even in identical apps, may still be overshadowed by the character of the transactions. Show us the top SQL in the statspack, too.

A bit of a reach for 4 times redo, but could it be you have delayed block cleanout? Perhaps one configuration has more batch update type jobs running? Different configuration on system jobs like statistics collection? One system is generating dump jobs or has more exports? (See the Let's Pretend post on Jonathan Lewis' blog).

Log Miner would let you see exactly what is going on.

Here's an odd one: http://www.freelists.org/post/oracle-l/Oracle-recommends-rebuilding-IOTs-in-AQ-to-reduce-redo,2

jg

--
_at_home.com is bogus.
http://www.signonsandiego.com/news/2009/dec/27/name-decade-it-was-mighty-mighty-bad/
Received on Mon Dec 28 2009 - 12:23:08 CST

Original text of this message