Re: Relocating undo tablespace

From: Mladen Gogala <gogala_at_sbcglobal.net>
Date: Thu, 09 Mar 2006 03:20:54 GMT
Message-ID: <pan.2006.03.09.03.20.53.794356@sbcglobal.net>

On Wed, 08 Mar 2006 07:32:12 -0800, EdStevens wrote:

>
> Herod wrote:

>> Your kidding right?
>>
>> That is the kind of thing statspack is exactly for.
>>
>> And as long as the database is brought up properly, only one copy of
>> the file is correct - EdStevens is correct
>> as well as the rest of us that said to do that :)

>
> In Mladen's defense, I'm not sure I see how statspack will show
> anything about where the SA placed a file system. What it will show is
> poor performance. It can even point that the poor performance is due
> to too many xxx_write wait events. That MAY (or may not) be due to the
> configuration of the disk farm. I don't think statspack can
> distinguish a db_file_parallel_write on a RAID-5 system from that on a
> SAME system.

What statspack will show is the same thing that V$FILESTAT will show: any

change in average I/O time and maximum I/O read and write times. That is
the only relevant thing here. No need to run statspack for that. Also,
using "sar -d" is advisable as well as "sar -b" and "iostat -x". If you

see any decrease in disk throughput, you should warn management of the possible performance loss.

-- 
http://www.mgogala.com

Received on Wed Mar 08 2006 - 21:20:54 CST