From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
cc: Robert Haas <robertmhaas@gmail.com>,
 Thomas Munro <thomas.munro@gmail.com>,
        Alvaro Herrera <alvherre@2ndquadrant.com>,
        Daniel Gustafsson <daniel@yesql.se>,
        "Jonathan S. Katz" <jkatz@postgresql.org>,
        Alexander Korotkov <a.korotkov@postgrespro.ru>,
        PostgreSQL Hackers <pgsql-hackers@postgresql.org>,
        Peter Geoghegan <pg@bowt.ie>, Bruce Momjian <bruce@momjian.us>,
        David Rowley <dgrowleyml@gmail.com>
Subject: Re: Improving connection scalability: GetSnapshotData()
In-reply-to: <1006917.1597602624@sss.pgh.pa.us>
References: <20200716012532.lfzpvg3w55d5rqrw@alap3.anarazel.de>
 <20200716013306.GA13861@alvherre.pgsql>
 <20200724011143.jccsyvsvymuiqfxu@alap3.anarazel.de>
 <CA+hUKG+Jpj7iMin_URBcyUoWxvkVjvQAd+LbUQ4NR4aZALUhjg@mail.gmail.com>
 <CA+hUKGK_4Sa0zHcfc+7HguHhQZ0dQn4P6yzMA1MsjHawmOq0ug@mail.gmail.com>
 <20200812001938.ixenvxdpg4ielbue@alap3.anarazel.de>
 <CA+hUKGKdey77=asfyBd+-Hnmz6Y_R+TZ+wtogaTppEdVTCH+rw@mail.gmail.com>
 <20200812173855.s4co5jz67z7tlnl7@alap3.anarazel.de>
 <872191.1597504251@sss.pgh.pa.us>
 <20200815164200.drok4tt5ju7ijivf@alap3.anarazel.de>
 <20200816181604.l54m6kss5ntd6xow@alap3.anarazel.de>
 <1006917.1597602624@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Sun, 16 Aug 2020 14:30:24 -0400"
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="----- =_aaaaaaaaaa0"
Content-ID: <1013229.1597609003.0@sss.pgh.pa.us>
Date: Sun, 16 Aug 2020 16:17:23 -0400
Message-ID: <1013484.1597609043@sss.pgh.pa.us>
Precedence: bulk

------- =_aaaaaaaaaa0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1013229.1597609003.1@sss.pgh.pa.us>

I wrote:
> It seems entirely likely that there's a timing component in this, for
> instance autovacuum coming along at just the right time.

D'oh.  The attached seems to make it 100% reproducible.

			regards, tom lane


------- =_aaaaaaaaaa0
Content-Type: text/x-diff; name="add-delay-before-vacuum.patch";
	charset="us-ascii"
Content-ID: <1013229.1597609003.2@sss.pgh.pa.us>
Content-Description: add-delay-before-vacuum.patch
Content-Transfer-Encoding: quoted-printable

diff --git a/src/test/isolation/specs/freeze-the-dead.spec b/src/test/isol=
ation/specs/freeze-the-dead.spec
index 915bf15b92..4100d9fc6f 100644
--- a/src/test/isolation/specs/freeze-the-dead.spec
+++ b/src/test/isolation/specs/freeze-the-dead.spec
@@ -32,6 +32,7 @@ session "s2"
 step "s2_begin"		{ BEGIN; }
 step "s2_key_share"	{ SELECT id FROM tab_freeze WHERE id =3D 3 FOR KEY SH=
ARE; }
 step "s2_commit"	{ COMMIT; }
+step "s2_wait"		{ select pg_sleep(60); }
 step "s2_vacuum"	{ VACUUM FREEZE tab_freeze; }
 =

 session "s3"
@@ -49,6 +50,7 @@ permutation "s1_begin" "s2_begin" "s3_begin" # start tra=
nsactions
    "s1_update" "s2_key_share" "s3_key_share" # have xmax be a multi with =
an updater, updater being oldest xid
    "s1_update" # create additional row version that has multis
    "s1_commit" "s2_commit" # commit both updater and share locker
+   "s2_wait"
    "s2_vacuum" # due to bug in freezing logic, we used to *not* prune upd=
ated row, and then froze it
    "s1_selectone" # if hot chain is broken, the row can't be found via in=
dex scan
    "s3_commit" # commit remaining open xact

------- =_aaaaaaaaaa0--