On Mon, 12 Nov 2001, Jeff Boes wrote:
> In article <20011109145054.H59285-100000@megazone23.bigpanda.com>,
> "Stephan Szabo" <sszabo@megazone23.bigpanda.com> wrote:
>
> > How many rows are in the table? Have you run vacuum analyze?
>
> Sorry, that information was in the original post, but perhaps you missed
> it:
>
> In article <9shhnf$23ks$1@news.tht.net>, "Jeff Boes" <jboes@nexcerpt.com>
> wrote:
>
> > We have a table which has approximately 400,000 rows. It has 17 columns,
> > and 4 indexes. The primary key is a int4 (filled by a sequence),
> > additionally we have two more int4 indexes and a timestamp index.
>
> Yes, VACUUM ANALYZE gets run every 24 hours, and currently the table
> grows by some 25K-40K rows per day. Could a factor be the time elapsed
> between the VACUUM and the query?
Is the 40000 row estimate for the number selected correct? If so, then
index scan may very well be a losing plan for this query. Does the forced
index scan actually take less time than the the sequence scan?