Re: PeopleSoft query
Date: Wed, 15 Jul 2009 02:13:07 +0800
Message-ID: <12ee65600907141113p2798f349o87142b4a260b454b_at_mail.gmail.com>
May I know what's good about OPTIMIZER_DYNAMIC_SAMPLING = 4? and the effects of this on the *WRK tables
_GBY_HASH_AGGREGATION_ENABLED, regarding this parameter.. yes I experienced a problem with this that made the reports of one of our clients inconsistent. They are on 10.2.0.3 RAC. The option is to upgrade to a newer patchset or apply this hidden parameter.
"I would always edit the dms scripts delivered by Oracle - that is the
source - and then reload the statement"
-> How do you reload the statement once you have edited the DMS script
and put some hints?
"When you get to a new version (or get a patch) do a flat file
comparison of the new and old dms scripts."
-> How do you make comparisons on the old and new scripts? Do you use
some kind of diff utility to compare the lines of the new and old?
"This is an either or choice with GP. Either you have the program analyze
the tables at runtime, or you remove and lock the stats and use ODS. If you
do both GPPDPRUN will crash because the analyze command will error."
-> Which is better? The problem we are having now is that the run
times are inconsistent that optimizer really goes bad sometimes. How
do you compute stats on SYSADM tables? how about indexes? and how
often you rebuild? how do you collect histograms?
(sorry about the self-publicity here, but chapter 11 of my book
discusses stored statements).
Yup, it's in the "to purchase" plan.
-> Actually, I got the ebook just earlier, lots of good stuff. I've
skimmed it earlier, but I'd like to have some more details on how to
maintain optimizer stats on a PeopleSoft environment. We all want
stable runtimes and the optimizer is the biggest factor :)