Date: Mon, 22 Jun 2020 15:27:32 -0400
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Kenneth Marshall <ktm@rice.edu>, pgsql-performance@lists.postgresql.org
Subject: Re: PostgreSQL 12.3 slow index scan chosen
Message-ID: <20200622192732.GA29072@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <1815407.1592677323@sss.pgh.pa.us>
User-Agent: Mutt/1.10.1 (2018-07-13)
Precedence: bulk

On 2020-Jun-20, Tom Lane wrote:

> I wrote:
> > ... oh, now I see: apparently, your filter condition is such that *no*
> > rows of the objectcustomfieldvalues table get past the filter:
> >
> >               ->  Index Scan using objectcustomfieldvalues3 on objectcustomfieldvalues objectcustomfieldvalues_1  (cost=0.56..807603.40 rows=915 width=4) (actual time=21165.441..21165.441 rows=0 loops=1)
> >                      Filter: ((disabled = 0) AND ((largecontent ~~* '%958575%'::text) OR ((content)::text ~~* '%958575%'::text)))
> >                      Rows Removed by Filter: 19030904

> You said you'd increased the stats target for
> objectcustomfieldvalues.objectid, but maybe the real problem is needing
> to increase the targets for content and largecontent, in hopes of driving
> down the estimate for how many rows will pass this filter condition.

... but those on content and largecontent are unanchored conditions --
are we still able to do any cardinality analysis using those?  I thought
not.  Maybe a trigram search would help?  See contrib/pg_trgm -- as far
as I remember that module is able to work with LIKE conditions.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services