From: Tom Lane <tgl@sss.pgh.pa.us>
To: Jeff Davis <pgsql@j-davis.com>
cc: Peter Geoghegan <pg@bowt.ie>, Alvaro Herrera <alvherre@2ndquadrant.com>,
        David Rowley <dgrowleyml@gmail.com>,
        Stephen Frost <sfrost@snowman.net>,
 Andres Freund <andres@anarazel.de>,
        Bruce Momjian <bruce@momjian.us>, Robert Haas <robertmhaas@gmail.com>,
        Justin Pryzby <pryzby@telsasoft.com>,
        Melanie Plageman <melanieplageman@gmail.com>,
        Tomas Vondra <tomas.vondra@2ndquadrant.com>,
        "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Subject: Re: Default setting for enable_hashagg_disk
In-reply-to: <571473265c5d5309f4012855ebd2131c136073ab.camel@j-davis.com>
References: <5cf74833a830e76cc7cc7506aa77bd6ab0c94f4c.camel@j-davis.com>
 <20200713164736.GA25726@alvherre.pgsql>
 <CAH2-WzmaJG4QYHq5LR+ZXEf1JMVcTzT9pEOeMFg0uuGXYDvgzQ@mail.gmail.com>
 <9e4a4cffe3eee0543e5b94c90821d39e03b81d29.camel@j-davis.com>
 <CAH2-Wz=qBF1QyXKdB6YNV993WpQ0wj-x4h8r9yjYUwQRCMkLjA@mail.gmail.com>
 <6d0aba397625df79bdc3ca85adc11b9e48ecb0da.camel@j-davis.com>
 <4118680.1595097043@sss.pgh.pa.us>
 <571473265c5d5309f4012855ebd2131c136073ab.camel@j-davis.com>
Comments: In-reply-to Jeff Davis <pgsql@j-davis.com>
	message dated "Sat, 18 Jul 2020 15:04:52 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <32870.1595121333.1@sss.pgh.pa.us>
Date: Sat, 18 Jul 2020 21:15:33 -0400
Message-ID: <32871.1595121333@sss.pgh.pa.us>
Precedence: bulk

Jeff Davis <pgsql@j-davis.com> writes:
> What is your opinion about pessimizing the HashAgg disk costs (not
> affecting HashAgg plans expected to stay in memory)? Tomas Vondra
> presented some evidence that Sort had some better IO patterns in some
> cases that weren't easily reflected in a principled way in the cost
> model.

Hm, was that in some other thread?  I didn't find any such info
in a quick look through this one.

> That would lessen the number of changed plans, but we could easily
> remove the pessimization without controversy later if it turned out to
> be unnecessary, or if we further optimize HashAgg IO.

Trying to improve our cost models under-the-hood seems like a
perfectly reasonable activity to me.

			regards, tom lane