Bruno Wolff III <bruno@wolff.to> writes:
> I have found that at least one query I do gets a poor plan after I update
> from CVS even if I don't need to do an initdb. Sometimes I have done an
> initdb to clear things up; most recently a vacuum analyze did the trick.
> This is a database used read only for web pages that I occasionally
> reload (part of the reload process is to do a vacuum analyze).
> Is this something I should expect? I would think since the stats were
> stored in the database, they would continue to be valid after updating
> the code (unless an initdb was forced).
If you didn't do initdb then I'd not expect pg_statistic to get wiped.
Details please?
regards, tom lane