RE: Peoplesoft Batch run and excessive redo generation
Date: Wed, 20 Oct 2010 08:48:39 -0400
Welcome to the world of PeopleSoft. The reason for the "excessive" redo generation has to do with what PeopleSoft calls a "temporary table". In fact these are real tables that get loaded, updated in some cases, unloaded, and repeat many times. In some cases the data in the temp table is needed for multiple sessions that run their data manipulation & report generator process (darn, can't for the life of me remember what that product's name is). Sometimes that is not true and you can replace it with a true Oracle temporary table to reduce your redo generation. Just don't expect any help from PeopleSoft on determining which is which. I did it by trial & error in a dev environment and then got the local developers and management to buy in, cut out half of the redo. BUT, WARNING: you are now moving into an area where PeopleSoft support can get dicey and upgrades a bit more of a pain, so document everything very well and create scripts to not only create what you need, but undo it as well.
Now for the required disclaimer: My experience with PeopleSoft is about 4 years old, so support which was bad at best may have gotten better with Oracle's acquisition of the product (bad move IMHO). So move cautiously, with much coordination, and be ready to back changes out.
Senior Oracle DBA
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Karl Arao Sent: Wednesday, October 20, 2010 1:50 AM To: oracle-l_at_freelists.org
Subject: Peoplesoft Batch run and excessive redo generation
We have this client with 30-50GB database on Peoplesoft.. and amazingly when it does batch processing it produces archivelogs (when compounded) way bigger than the database.. this is around 150GB worth of archivelogs.
The issue here is on the Disaster Recovery, if we turn on the log transfer of Data Guard on the period of the batch run the network bandwidth capacity gets consumed easily causing long queue for archive gap. This is the same if we turn off the log transfer, then do the batch run, then re-enable it... the gap resolution process will eat up the entire network bandwidth.
I have read on the doc "Batch Processing in Disaster Recovery Configurations - Best Practices for Oracle Data Guard" https://docs.google.com/viewer?url=http://www.hitachi.co.jp/Prod/comp/so ft1/oracle/pdf/OBtecinfo-08-008.pdf but this is 11g, on 10g what I can do is SSH compression http://goo.gl/jc6n but still even with compression enabled the redo requirement may still exceed the network capacity
The RMAN incremental backups (differential & cumulative) does not help here because it will generate and re-apply the same amount of redo
Another thing I'm exploring is just turn off the log transport before doing the batch run, then after that.. duplicate the standby (not incremental update but fresh)
Do you have other ideas around this? If you have encountered this how did you address the issue? Is there a related PeopleSoft bug? This will be very much appreciated.
-- Karl Arao karlarao.wordpress.com karlarao.tiddlyspot.com -- http://www.freelists.org/webpage/oracle-lReceived on Wed Oct 20 2010 - 07:48:39 CDT