MIME-Version: 1.0
References: <1959506.1775527693@sss.pgh.pa.us>
 <CABPTF7WPRVJGdDeuWo-=3csnDs5FGfUsg97xppPyzoj9fRAMeQ@mail.gmail.com>
 <CABPTF7WAcV_uEaJa3jmYPc26rApa64hV5HmjRBN=OS17MAybUg@mail.gmail.com>
 <CAPpHfdvM9Tp1_g1JY9H+cx1H0Q4jH-BZFZGQwgviLNs1HSg+CA@mail.gmail.com>
 <CABPTF7X0iV=kGC4gjsTj4NvK_NNEJGM3YTc7Obxs5GOiYoMhEw@mail.gmail.com>
 <CABPTF7UBdEfyxATWntmCfoJrwB6iPrnhkXO7y_Avmqc2bOn27A@mail.gmail.com>
 <CAPpHfduhQsm44j_ziZ6ykFTDZ2SFvZp04iCrc5_dD9PGZUrhrw@mail.gmail.com>
 <d2ogsp54if47w35kx4vu7o3nkfghpuzctwbaudzey5brxavomw@bgfs3p7twoib>
 <CABPTF7W7=uK0ypteDLRTiR4PnRP-BEUEJJtit+Q17aiPPhpLrw@mail.gmail.com>
 <CABPTF7Ukk8iJF7TpnK2mFOaboNJgWL1csfXu4e3J4GT0o7x0GQ@mail.gmail.com>
 <CAPpHfdupZ3ooxmnJvhjx7rnZ7VMcz_uyxGC4drMSEM_P3PtGzw@mail.gmail.com>
 <CABPTF7WJ35p7uidJJZs7fzxBtbVL_0xSFUdZ2Fe8pXh00e=Mxw@mail.gmail.com>
 <CAPpHfdufWG032J=fyv1eWoveeyPwqJ57PGU2edA5OsOmexGDTw@mail.gmail.com>
 <CABPTF7W-gaO=FAkhda=_pDQJjLne68ioNHHU8vuB4iEnswR1=w@mail.gmail.com>
 <e4e98e40-40d5-4d53-85de-6c456c2b301a@iki.fi>
 <CABPTF7XV=rPe9iVTyR5O=CT0oMUYTPJpr+4+T35n0oKeaaNZHQ@mail.gmail.com>
 <CABPTF7V1GqXRe9Kyiw=oyDoUU4Oss49nMBhco12V7xb72P6WvQ@mail.gmail.com>
 <CABPTF7XdpM8LCvrkQdXrj3qt88auofiVw_STbUnXg0GEuzbDNw@mail.gmail.com>
In-Reply-To: 
 <CABPTF7XdpM8LCvrkQdXrj3qt88auofiVw_STbUnXg0GEuzbDNw@mail.gmail.com>
From: Alexander Korotkov <aekorotkov@gmail.com>
Date: Wed, 8 Jul 2026 20:38:28 +0300
Message-ID: 
 <CAPpHfdvu7bPzP1p3eZW3t6QgF39stTce=OyObyn6N2J+ZC5Khw@mail.gmail.com>
Subject: Re: Implement waiting for wal lsn replay: reloaded
To: Xuneng Zhou <xunengzhou@gmail.com>
Cc: Heikki Linnakangas <hlinnaka@iki.fi>, Andres Freund <andres@anarazel.de>,
 Tom Lane <tgl@sss.pgh.pa.us>,
	Peter Eisentraut <peter@eisentraut.org>,
 Thomas Munro <thomas.munro@gmail.com>,
	=?UTF-8?Q?=C3=81lvaro_Herrera?= <alvherre@kurilemu.de>,
	Chao Li <li.evan.chao@gmail.com>,
 pgsql-hackers <pgsql-hackers@lists.postgresql.org>,
	Michael Paquier <michael@paquier.xyz>, jian he <jian.universality@gmail.com>,
	Tomas Vondra <tomas@vondra.me>, Yura Sokolov <y.sokolov@postgrespro.ru>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: 
 <https://www.postgresql.org/message-id/CAPpHfdvu7bPzP1p3eZW3t6QgF39stTce%3DOyObyn6N2J%2BZC5Khw%40mail.gmail.com>
Precedence: bulk

On Wed, Jul 8, 2026 at 3:08=E2=80=AFPM Xuneng Zhou <xunengzhou@gmail.com> w=
rote:
> On Mon, Jul 6, 2026 at 10:17=E2=80=AFPM Xuneng Zhou <xunengzhou@gmail.com=
> wrote:
> >
> > On Mon, Jul 6, 2026 at 9:49=E2=80=AFPM Xuneng Zhou <xunengzhou@gmail.co=
m> wrote:
> > >
> > > Hi Heikki,
> > >
> > > Thanks for looking into this!
> > >
> > > On Mon, Jul 6, 2026 at 7:04=E2=80=AFPM Heikki Linnakangas <hlinnaka@i=
ki.fi> wrote:
> > >                       /*
> > > > >                        * Apply the record
> > > > >                        */
> > > > >                       ApplyWalRecord(xlogreader, record, &replayT=
LI);
> > > > >
> > > > >                       /*
> > > > >                        * Wake up processes waiting for standby re=
play, write, or flush
> > > > >                        * LSN to reach current replay position.  R=
eplay implies that the
> > > > >                        * WAL was already written and flushed to d=
isk, so write and flush
> > > > >                        * waiters can be woken at the replay posit=
ion too.
> > > > >                        */
> > > > >                       WaitLSNWakeup(WAIT_LSN_TYPE_STANDBY_REPLAY,
> > > > >                                                 XLogRecoveryCtl->=
lastReplayedEndRecPtr);
> > > > >                       WaitLSNWakeup(WAIT_LSN_TYPE_STANDBY_WRITE,
> > > > >                                                 XLogRecoveryCtl->=
lastReplayedEndRecPtr);
> > > > >                       WaitLSNWakeup(WAIT_LSN_TYPE_STANDBY_FLUSH,
> > > > >                                                 XLogRecoveryCtl->=
lastReplayedEndRecPtr);
> > > >
> > > > That's not wrong, but I've got some comments:
> > > >
> > > > 1. It's reading XLogRecoveryCtl->lastReplayedEndRecPtr without a lo=
ck or
> > > > atomics. That's ok, no other process modifies lastReplayedEndRecPtr=
, but
> > > > it feels a little dirty.
> > > >
> > > > 2. We're now doing three extra function calls on every WAL record. =
This
> > > > is a very hot path, and most of the time, we'll just take the fast =
path
> > > > in WaitLSNWakeup to return without doing anything. Andres and other=
s
> > > > assumed up-thread that it's negligible (we used to have pre-checks =
here
> > > > in the caller), but I wonder if you did any performance testing?
> > >
> > > Agreed, this is a hot path. The performance impact of these extra
> > > calls doing real work hasn't been measured yet. I'll do some testing.
> > >
> > > > 3. There are other "wakeup" calls inside ApplyWalRecord(), to wake =
up
> > > > walsenders and walreceivers. They could perhaps use the same wait-l=
sn
> > > > machinery now, but that's v20 material. However, I think these
> > > > WaitLSNWakeup() calls should also be moved inside ApplyWalRecord(),=
 so
> > > > that we'd have all the wakeup actions in one place.
> > >
> > > + 1. This makes the code safer and more readable.
> > >
> > > > 4. Once you move those calls inside ApplyWalRecord(), like this:
> > > >
> > > > > @@ -1979,20 +1979,30 @@ ApplyWalRecord(XLogReaderState *xlogreade=
r, XLogRecord *record, TimeLineID *repl
> > > > >         /*
> > > > >          * Update lastReplayedEndRecPtr after this record has bee=
n successfully
> > > > >          * replayed.
> > > > >          */
> > > > >         SpinLockAcquire(&XLogRecoveryCtl->info_lck);
> > > > >         XLogRecoveryCtl->lastReplayedReadRecPtr =3D xlogreader->R=
eadRecPtr;
> > > > >         XLogRecoveryCtl->lastReplayedEndRecPtr =3D xlogreader->En=
dRecPtr;
> > > > >         XLogRecoveryCtl->lastReplayedTLI =3D *replayTLI;
> > > > >         SpinLockRelease(&XLogRecoveryCtl->info_lck);
> > > > >
> > > > > +       /*
> > > > > +        * Wake up processes waiting for standby replay, write, o=
r flush LSN to
> > > > > +        * reach current replay position.  Replay implies that th=
e WAL was already
> > > > > +        * written and flushed to disk, so write and flush waiter=
s can be woken at
> > > > > +        * the replay position too.
> > > > > +        */
> > > > > +       WaitLSNWakeup(WAIT_LSN_TYPE_STANDBY_REPLAY, xlogreader->E=
ndRecPtr);
> > > > > +       WaitLSNWakeup(WAIT_LSN_TYPE_STANDBY_WRITE, xlogreader->En=
dRecPtr);
> > > > > +       WaitLSNWakeup(WAIT_LSN_TYPE_STANDBY_FLUSH, xlogreader->En=
dRecPtr);
> > > > > +
> > > > >         /* ------
> > > > >          * Wakeup walsenders:
> > > > >          *
> > > > >          * On the standby, the WAL is flushed first (which will o=
nly wake up
> > > > >          * physical walsenders) and then applied, which will only=
 wake up logical
> > > > >          * walsenders.
> > > >
> > > > It becomes clear that you don't actually need the memory barrier in=
side
> > > > WaitLSNWakeup(). Not sure if they're needed for other callers, but =
here
> > > > we have just released a spinlock, which acts as a memory barrier. I=
t
> > > > might not be worth relaxing, but it does seem a little silly.
> > >
> > > If we made the move here, I think the memory barrier could be relaxed
> > > since other callers are guarded by either the spinlock or full-barrie=
r
> > > atomic write already.  We might also want to make the contract of
> >
> > OK, the 'if' here is redundant...
>
> After revisiting the memory barrier in WaitLSNWakeup and why it is
> introduced there in a80a593ab63 rather than recalling it from memory,
> I think relaxing it here could be unsafe.
>
> In WaitLSNWakeup(), use pg_atomic_read_membarrier_u64() in the
> fast-path check so the waker's preceding position store is globally
> visible before minWaitedLSN is read.
>
> Without the barrier in WaitLSNWakeup(), this interleaving is possible:
>
> Initial:
>   minWaitedLSN =3D PG_UINT64_MAX
>   replayLSN =3D 90
>
> Waiter:
>   stores minWaitedLSN =3D 100
>   reads replayLSN before the waker publishes the new replay position
>   sees replayLSN =3D 90
>   decides it should sleep
>
> Waker:
>   publishes replayLSN =3D 100
>   reads old minWaitedLSN =3D PG_UINT64_MAX
>   skips the wakeup
>
> Then the waiter goes to sleep even though replay has reached its
> target LSN. To avoid this, we still need to make sure that the
> publication of replayLSN precedes the read of minWaitedLSN, so that
> the waker cannot decide "nobody is waiting" before its own progress is
> still not visible to the waiter.

Yes, I also think the memory barrier for waker between publishing
replayLSN and reading minWaitedLSN is required.  However, sequence of
three WaitLSNWakeup() calls makes 3 memory barriers while only one is
required.  We could introduce a hierarchy for WAIT_LSN_TYPE_STANDBY_*:
WAIT_LSN_TYPE_STANDBY_FLUSH implies WAIT_LSN_TYPE_STANDBY_WRITE,
WAIT_LSN_TYPE_STANDBY_REPLAY implies WAIT_LSN_TYPE_STANDBY_WRITE and
WAIT_LSN_TYPE_STANDBY_FLUSH.  Then ApplyWalRecord() can call
WaitLSNWakeup() only once and make only 1 memory barrier.

------
Regards,
Alexander Korotkov
Supabase