MIME-Version: 1.0
References: <7daef094-abf3-4672-bc23-3df4763b16a3@gmail.com>
 <aiP/X1FThlZRCFiN@bdtpg>
 <OS9PR01MB1214908BA67A7811BD6281208F51C2@OS9PR01MB12149.jpnprd01.prod.outlook.com>
 <CABPTF7WmLKOJcSPod7zT2hynUFQcUs0VoQyR-p=XFsafvmGm7g@mail.gmail.com>
In-Reply-To: 
 <CABPTF7WmLKOJcSPod7zT2hynUFQcUs0VoQyR-p=XFsafvmGm7g@mail.gmail.com>
From: Xuneng Zhou <xunengzhou@gmail.com>
Date: Mon, 8 Jun 2026 21:00:17 +0800
Message-ID: 
 <CABPTF7U_B9pdC563fFONQLX_FXCZtxZgM+jN9snzMVg9b9MRfg@mail.gmail.com>
Subject: Re: t/035_standby_logical_decoding.pl might fail on attempt to read
 wrong timeline
To: "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>
Cc: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>,
 Alexander Lakhin <exclusion@gmail.com>,
	pgsql-hackers <pgsql-hackers@postgresql.org>
Content-Type: multipart/mixed; boundary="0000000000006e5ea60653bd982c"
Archived-At: 
 <https://www.postgresql.org/message-id/CABPTF7U_B9pdC563fFONQLX_FXCZtxZgM%2BjN9snzMVg9b9MRfg%40mail.gmail.com>
Precedence: bulk

--0000000000006e5ea60653bd982c
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Mon, Jun 8, 2026 at 4:31=E2=80=AFPM Xuneng Zhou <xunengzhou@gmail.com> w=
rote:
>
> Hi Kuroda-san,
>
> On Mon, Jun 8, 2026 at 12:25=E2=80=AFPM Hayato Kuroda (Fujitsu)
> <kuroda.hayato@fujitsu.com> wrote:
> >
> > Hi Alexander, Bertrand, Xuneng,
> >
> > Thanks for seeing the failure. Our team also recognized but could not f=
ind the reason.
> >
> > > Yeah, it looks like there is a race condition here. I think we should=
 check if
> > > the insertion timeline has already been set (like the walsummarizer i=
s doing).
> >
> > Sorry for stupid question; I tried to reproduce the failure but could n=
ot, see attached.
> >
> > IIUC, the issue can happen if the walsender must read the WAL record ge=
nerated
> > after the promotion but the timeline could not be updated.
>
> I think the race is that the logical walsender is trying to read the
> pre-promotion wal from the slot's restart_lsn.
>
> 1) In 035_standby_logical_decoding.pl, the active slot /
> pg_recvlogical process is started before:
> INSERT rows 1..4   # before promotion
> promote standby
> INSERT rows 5..7   # after promotion
>
> 2) Alexander=E2=80=99s reproducer sleeps immediately after:
> ReplicationSlotAcquire(cmd->slotname, true, true);
> pg_usleep(200000);
>
> So with the delay:
> slot is active but walsender has not yet create decoding context and
> called xlogbeginread();
>
> 3) Then the test continues and promotes the standby while the
> walsender still sleeping, therefore promotion could happen before the
> walsender starts its actual wal read.
>
> 4)  The logical walsender is trying to read the pre-promotion wal from
> the slot's restart_lsn. The failing log implies this:
> Streaming transactions committing after 0/06487A20,
> reading WAL from 0/064879E8.
>
> This line is emitted from:
> errdetail("Streaming transactions committing after %X/%08X, reading
> WAL from %X/%08X.",
>           slot->data.confirmed_flush,
>           slot->data.restart_lsn)
>
> Function StartLogicalReplication() initiates the xlogreader at the restar=
t lsn:
> XLogBeginRead(logical_decoding_ctx->reader,
>               MyReplicationSlot->data.restart_lsn);
>
> Because the slot was created and acquired before promotion, its
> restart_lsn is from the standby-era state which points to wal required
> for decoding the pre-promotion stream. In the test, the first expected
> output is rows 1..4, inserted before promotion.
>
> The failing output was empty:
> got: ''
> expected:
> BEGIN
> rows 1..4
> COMMIT
> BEGIN
> rows 5..7
> COMMIT
>
> Empty output suggests that the walsender died before emitting even the
> first pre-promotion transaction.
>
> 5) The walsender try to read wal but failed
> Failed run:
>  LOG:  !!!WalSndSegmentOpen| nextSegNo: 6,
> path: pg_wal/000000010000000000000006
> ERROR:  requested WAL segment
> 000000010000000000000006 has already been removed
>
> Passed run
> LOG:  !!!WalSndSegmentOpen| nextSegNo: 6,
> path: pg_wal/000000020000000000000006
>
> As the log suggested, the walsender of the successful run readed the
> wal from timeline 2 but failed to do so for timeline1.
> This comment in xlogutils.is almost exactly the issue:
>
> We care about timelines ... reading xlog generated prior to a promotion
> ...
> The server copied the segment to the new timeline ... there's no guarante=
e
> the old segment will still exist. It may have been deleted or renamed wit=
h
> a .partial suffix
>
> So the need for tli1 does not always imply that the wal was generated
> after promotion. It could be bytes from before the switchpoint, but
> they now be available through the new tli segment file.
>
> 6) XLogReadDetermineTimeline() needs the current system-wide TLI.
>
> The current code in does roughly:
>
>          am_cascading_walsender =3D RecoveryInProgress();
>
>          if (am_cascading_walsender)
>                  GetXLogReplayRecPtr(&currTLI);
>          else
>                  currTLI =3D GetWALInsertionTimeLine();
>
> During promotion, we set the wal insertion timeline before marking
> recovery as fully done. So there is a window where:
>
> - InsertTimeLineID is already the new timeline, e.g. tli 2.
> - RecoveryInProgress() can still return true in another backend.
> - A logical walsender that started before promotion can wake up and
> choose its read timeline using the old recovery-state test.
>
> In the race, replay tli 1 is chosen even though the promotion has
> created the new tli 2 wal segment copy. Then WalSndSegmentOpen() tries
> to open 000000010000000000000006, which may already be gone or
> renamed, while the correct file is 000000020000000000000006.
>
> 7) Bertrand's proposal to fix this
> Recognizing the intermediate promotion state -- if recovery still
> shows progress but the insertion tli is already advanced, use that
> insertion tli for the timeline to be chosen.
>
> > However, I think logical_read_xlog_page() is called after the new WAL r=
ecords
> > are generated, i.e., am_cascading_walsender has already been false at t=
hat time.
> > So not sure where is the race?
>
> This function is not only called after new post-promotion records are
> generated. When START_REPLICATION begins, CreateDecodingContext() /
> XLogBeginRead() starts from the slot's restart_lsn, which can be well
> before promotion.
>

I tweaked the reproducer based on the theory outlined above. The main
changes from the original reproducer are:

1) blocks at logical-walsender-after-slot-acquire in walsender.c,
before the decoding context is created and before the reader starts
from restart_lsn, matching the delay set by Alexander

2) Forces the first read to occur during promotion. It inserts rows
1..4, waits for replay, starts promotion with pg_promote(false), holds
startup at startup-logical-decoding-status-change-end-of-recovery,
then wakes the walsender.

The test failed on HEAD and passed after applying the patch by Bertrand.

--=20
Regards,
Xuneng Zhou
HighGo Software Co., Ltd.

--0000000000006e5ea60653bd982c
Content-Type: application/octet-stream;
	name="v1-0001-Reproducer-for-logical-decoding-choosing-the-wron.patch"
Content-Disposition: attachment;
	filename="v1-0001-Reproducer-for-logical-decoding-choosing-the-wron.patch"
Content-Transfer-Encoding: base64
Content-ID: <f_mq57xcb60>
X-Attachment-Id: f_mq57xcb60
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--0000000000006e5ea60653bd982c--