On Thu, Jan 03, 2008 at 07:11:07AM +0200, Brian Modra wrote:
> Thanks, I think you have me on the right track. I'm testing a vacuum
> analyse now to see how long it takes, and then I'll set it up to
> automatically run every night (so that it has a chance to complete
> before about 6am.)

Note that "VACUUM ANALYSE" and "ANALYSE" are not identical: the former also
performs vacuum.  On a table that is not updating that often but that is
expanding rapidly, you may not need that extra I/O.  Analyse on its own can
perform just the statistical sampling.  If you're not creating dead tuples
with UPDATE, DELETE, or ROLLBACK, that might be enough most of the time.

A

In response to

• Re: Index performance at 2008-01-03 05:11:07 from Brian Modra

pgsql-hackers by date

Next:	From: Kevin Grittner	Date: 2008-01-04 18:16:13
	Subject: OUTER JOIN performance regression remains in 8.3beta4
Previous:	From: Andrew Sullivan	Date: 2008-01-04 18:06:04
	Subject: Re: Dynamic Partitioning using Segment Visibility Maps