Re: redo copy latch

From: Guy Harrison <gharriso_at_werple.net.au>
Date: 1997/03/17
Message-ID: <01bc325a$ee1db120$0100007f@gharriso>#1/1

Randy,

LOG_ENTRY_PREBUILD_THRESHOLD has now become "undocumented" and is preceeded by an underscore: _LOG_ENTRY_PREBUILD_THRESHOLD.

I think your problem is that you've set LOG_SMALL_ENTRY_MAX_SIZE to 0. This causes all redo entries to be made on the redo allocation latch. If you want all copies to be made on the redo copy latch set LOG_SMALL_ENTRY_MAX_SIZE to 1. Note that this doesn't "eliminate" gets on redo allocation latch: it just reduces the amount of time the allocation latch is held.

Guy

-- 

----------------------------------------------------------------------
Guy Harrison								
gharriso_at_werple.net.au				
http://werple.net.au/~gharriso		
(61) 0419 377 964								


Randy H. Julien <julienr_at_ncr.disa.mil> wrote in article
<33285DD7.3031_at_ncr.disa.mil>...


> Fellow DBA's

> 

> I am trying to improve upon my very poor .003 redo copy latch hit ratio

> by prebuilding some/all of the redo entries.  Documentation states that

> setting init.ora entry LOG_ENTRY_PREBUILD_THRESHOLD to > 0 bytes, along

> with increased LOG_SIMULTANEOUS_COPIES, does the trick. Problem is, it

> appears that Oracle no longer supports the THRESHOLD setting.  I queried

> v$parameter and it doesn't appear. I am running Oracle 7.3.2.1 on

> SUN/Solaris 2.4 with 4 processors.  Has this setting been replaced.  And

> if so, how can I improve the nearly 0 hit ratio?  Note that the COPIES

> setting is at 8 (twice # of CPUs), and LOG_SMALL_ENTRY_MAX_SIZE = 0 to

> eliminate the redo allocation latch.

> 

> Any help is appreciated. Thanks!

> 

> Randy

>

Received on Mon Mar 17 1997 - 00:00:00 CST