Date: Thu, 27 Sep 2018 14:33:14 -0500
From: Justin Pryzby <pryzby@telsasoft.com>
To: Arjun Ranade <ranade@nodalexchange.com>
Cc: pgsql-performance@postgresql.org
Subject: Re: SELECT statement returns in 10seconds, but INSERT/CREATE TABLE
 AS with same SELECT takes 7 minutes
Message-ID: <20180927193313.GI776@telsasoft.com>
References: 
 <CANrrCRxsC3wr711bjAJ4zK_16vF-qQHZXjSJTstbQt9kG5QDSw@mail.gmail.com>
 <20180927185224.GF776@telsasoft.com>
 <CANrrCRy_HSsYCTDVH+EtDoBJzFiEob3Nd5bztmno5V13WFWOjQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <CANrrCRy_HSsYCTDVH+EtDoBJzFiEob3Nd5bztmno5V13WFWOjQ@mail.gmail.com>
User-Agent: Mutt/1.5.23 (2014-03-12)
Precedence: bulk

On Thu, Sep 27, 2018 at 03:37:57PM -0400, Arjun Ranade wrote:
> Yes, that join is concerning (red text below).  The conditions all need to
> be checked so they are independent.

You can play with the join conditions to see which test is getting such a bad
estimate, or if it's a combination of tests (as I suspected) giving a bad
estimate.

There's a good chance this one isn't doing very well:

>    vw2.product_group_name ||'.'|| vw2.product_node_name = i.product_node_name

As a workaround/test, you could maybe add an expression index
ON( (vw2.product_group_name ||'.'|| vw2.product_node_name) )

..and then ANALYZE.  Eventually, you'd want to consider splitting
i.product_node_name into separate columns.  

Justin