From: Gregory Stark <stark@enterprisedb.com>
To: "Henrik Zagerholm" <henke@mac.se>
Cc: <pgsql-performance@postgresql.org>,  <pgsql-general@postgresql.org>
Subject: Re: [PERFORM] Planner making wrong decisions 8.2.4. Insane cost
	calculations.
In-Reply-To: <0C963119-103A-4EB1-B9D8-0FD3E16A11F0@mac.se> (Henrik Zagerholm's
	message of "Mon, 6 Aug 2007 14:21:10 +0200")
Organization: EnterpriseDB
References: <0C963119-103A-4EB1-B9D8-0FD3E16A11F0@mac.se>
Date: Mon, 06 Aug 2007 14:07:17 +0100
Message-ID: <87abt4n75m.fsf@oxford.xeocode.com>
User-Agent: Gnus/5.110006 (No Gnus v0.6) Emacs/21.4 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii

"Henrik Zagerholm" <henke@mac.se> writes:

> Hi list,
>
> I'm having a weird acting query which simply retrieves some files stored in a db
> which are related to a specific archive and also has a  size lower than 1024
> bytes.
> Explain analyze below. The first one is with seq-scan enabled and the other one
> with seq-scans disabled. The weird thing is the seq scan on  tbl_file_structure
> and also the insane calculated cost of 100 000 000  on  some tables.

Well the way Postgres disables a plan node type is by giving it a cost of
100,000,000. What other way did you expect it to be able to scan
tbl_filetype_suffix anyways? What indexes do you have on tbl_filetype_suffix?

And any chance you could resend this stuff without the word-wrapping?
It's pretty hard to read like this:

"                                      ->  Seq Scan on tbl_filetype_suffix
(cost=100000000.00..100000001.34 rows=14  width=8) (actual time=0.133..0.176
rows=14 loops=1)"
"                                            Filter: (filetype_suffix_index IS
TRUE)"


-- 
  Gregory Stark
  EnterpriseDB          http://www.enterprisedb.com