Date: Fri, 18 Oct 2019 02:13:34 +0200
From: Tomas Vondra <tomas.vondra@2ndquadrant.com>
To: Michael Lewis <mlewis@entrata.com>
Cc: postgres performance list <pgsql-performance@postgresql.org>
Subject: Re: Reading explain plans- row estimates/actuals on lower nodes vs
 next level up
Message-ID: <20191018001334.uhjcp7kpbj46gl5e@development>
References: 
 <CAHOFxGo1UAF45F_fdZML1QKmSjjvN_npLx52yQv7NV6Xk=COJg@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Disposition: inline
In-Reply-To: 
 <CAHOFxGo1UAF45F_fdZML1QKmSjjvN_npLx52yQv7NV6Xk=COJg@mail.gmail.com>
User-Agent: NeoMutt/20180716
Precedence: bulk

On Thu, Oct 17, 2019 at 03:15:21PM -0600, Michael Lewis wrote:
>https://explain.depesz.com/s/Caa5
>
>I am looking at this explain analyze output and seeing a nested loop
>towards the lowest levels with pretty bad estimate vs actual (2.3k vs 99k),
>but the things that feed that nested loop seem like the estimates are
>rather close (index scans with 11 estimated vs 30 actual and 3350 vs
>3320)... why does the higher node have such a different estimate vs actual
>ratio?
>

This is usually a sign of non-uniform distribution for the join columns,
and/or correlation between sides of the join.

regards

-- 
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services