Re: Statistics contradiction can cause wrong cardinality estimation
Date: Tue, 22 Jul 2008 11:54:51 -0700
A couple comments:
I would recommend to always use YYYY as RRRR/RR was introduced as a
This should yield a number value for Pstart/Pstop vs. a KEY/KEY plan.
Based on your note that recollecting stats resolved the issue, what command was used to gather stats the produced the bad plans? What command was used to produce the good plans? What differences in the stats did you observe?