From: Tom Lane <tgl@sss.pgh.pa.us>
To: Nathan Bossart <nathandbossart@gmail.com>
cc: Robert Haas <robertmhaas@gmail.com>, Andres Freund <andres@anarazel.de>,
        pgsql-hackers@postgresql.org, schneider@ardentperf.com
Subject: Re: monitoring usage count distribution
In-reply-to: <20230406180608.GA275542@nathanxps13>
References: 
 <CA+Tgmoa-c3avHFBdko9d7bGQuYXrJhHHjhG1yAs8h-k_sLbyPQ@mail.gmail.com>
 <20230404232919.uibzbhjdylk3mlvp@awork3.anarazel.de>
 <CA+TgmobrQ90KGN4dwxL8q-LV529gCMmzXiy8CG5eQwjYTPzZLA@mail.gmail.com>
 <20230405175124.GA31516@nathanxps13>
 <CA+TgmoYRHDDJTgPcFkomz4KzzsQHXRJtqKeJyOk4AZNTjf=TnQ@mail.gmail.com>
 <20230405190921.ck7v4y2h3ogoanvg@awork3.anarazel.de>
 <20230405194101.GB105962@nathanxps13> <2061855.1680725784@sss.pgh.pa.us>
 <CA+TgmobRvYRxubzHY7PgnuYhJHPBGSF8a-6Ax+cz0Fm0EbxGgA@mail.gmail.com>
 <2478091.1680802355@sss.pgh.pa.us> <20230406180608.GA275542@nathanxps13>
Comments: In-reply-to Nathan Bossart <nathandbossart@gmail.com>
	message dated "Thu, 06 Apr 2023 11:06:08 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <3089321.1680892171.1@sss.pgh.pa.us>
Date: Fri, 07 Apr 2023 14:29:31 -0400
Message-ID: <3089322.1680892171@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/3089322.1680892171%40sss.pgh.pa.us>
Precedence: bulk

Nathan Bossart <nathandbossart@gmail.com> writes:
> On Thu, Apr 06, 2023 at 01:32:35PM -0400, Tom Lane wrote:
>> There seems to be enough support for the existing summary function
>> definition to leave it as-is; Andres likes it for one, and I'm not
>> excited about trying to persuade him he's wrong.  But a second
>> slightly-less-aggregated summary function is clearly useful as well.
>> So I'm now thinking that we do want the patch as-submitted.
>> (Caveat: I've not read the patch, just the description.)

> In case we want to do both, here's a 0002 that changes usagecount_avg to an
> array of usage counts.

I'm not sure if there is consensus for 0002, but I reviewed and pushed
0001.  I made one non-cosmetic change: it no longer skips invalid
buffers.  Otherwise, the row for usage count 0 would be pretty useless.
Also it seemed to me that sum(buffers) ought to agree with the
shared_buffers setting.

			regards, tom lane