MIME-Version: 1.0
References: <c20f6340eb26f3b736abc59471bfada8@oss.nttdata.com>
 <CAGXjcj=hnEZCCDDMRv06EQnDcv0mRe6+Ph1gv8=Hb7NEc51y5A@mail.gmail.com>
 <CAM6-o=DnHSj2q9_+Wohrvagd-QAqvpoepNRS9LRadMBCkTAMkw@mail.gmail.com>
 <CAGECzQSOrmOTSyVHi1tkx9QDg9_PkVJudtdTH0B7ovHeZXtFLQ@mail.gmail.com>
In-Reply-To: 
 <CAGECzQSOrmOTSyVHi1tkx9QDg9_PkVJudtdTH0B7ovHeZXtFLQ@mail.gmail.com>
From: Atsushi Torikoshi <torikoshia.tech@gmail.com>
Date: Mon, 25 May 2026 21:17:56 +0900
Message-ID: 
 <CAM6-o=BL59LgruKXm3DCjzFzvr7TTJPtryEPj51YeGMFrWO0EQ@mail.gmail.com>
Subject: Re: RFC: Allow EXPLAIN to Output Page Fault Information
To: Jelte Fennema-Nio <postgres@jeltef.nl>
Cc: torikoshia <torikoshia@oss.nttdata.com>,
	Pgsql Hackers <pgsql-hackers@postgresql.org>, Tomas Vondra <tomas@vondra.me>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: 
 <https://www.postgresql.org/message-id/CAM6-o%3DBL59LgruKXm3DCjzFzvr7TTJPtryEPj51YeGMFrWO0EQ%40mail.gmail.com>
Precedence: bulk

On Sun, May 24, 2026 at 6:50=E2=80=AFPM Jelte Fennema-Nio <postgres@jeltef.=
nl> wrote:
>
> On Fri, 22 May 2026 at 18:07, Atsushi Torikoshi
> <torikoshia.tech@gmail.com> wrote:
> > Updated the patch to clarify that the reported values include not only
> > the backend process executing the query, but also any parallel query
> > worker processes involved in the query.
>
> For PG19 Tomas added an IO option to EXPLAIN. I think it would be
> better to have the Storage IO tracking be enabled by the IO option,
> rather than the BUFFERS option.

Thank you for pointing that out.
I agree that having this under the IO option seems more intuitive.

One thing I am slightly concerned about, though, is that according to
the documentation, the IO option reports only I/O performed by scan
nodes:

> Include information on I/O performed by scan nodes providing such informa=
tion.

In contrast, the information produced by this patch is based on
getrusage(), so it includes all Storage I/O performed by the backend
and parallel workers, not just scan nodes.
For example, it would also include temporary file I/O generated by
operations such as sorts that spill to disk when they exceed work_mem.

I suppose this could be clarified in the documentation, but I would be
interested to hear if anyone has concerns or objections about that
approach.

--
Regards,
Atsushi Torikoshi