Date: Mon, 23 Aug 2021 13:14:09 -0400
From: "alvherre@alvh.no-ip.org" <alvherre@alvh.no-ip.org>
To: "Bossart, Nathan" <bossartn@amazon.com>
Cc: Robert Haas <robertmhaas@gmail.com>,
	Kyotaro Horiguchi <horikyota.ntt@gmail.com>,
	"x4mmm@yandex-team.ru" <x4mmm@yandex-team.ru>,
	"a.lubennikova@postgrespro.ru" <a.lubennikova@postgrespro.ru>,
	"hlinnaka@iki.fi" <hlinnaka@iki.fi>,
	"matsumura.ryo@fujitsu.com" <matsumura.ryo@fujitsu.com>,
	"masao.fujii@gmail.com" <masao.fujii@gmail.com>,
	"pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Subject: Re: archive status ".ready" files may be created too early
Message-ID: <202108231714.kdyudmc5aojp@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <4268438D-343C-4831-8511-17B03CEA028E@amazon.com>
Archived-At: 
 <https://www.postgresql.org/message-id/202108231714.kdyudmc5aojp%40alvherre.pgsql>
Precedence: bulk

On 2021-Aug-23, Bossart, Nathan wrote:

> Sorry, I'm still not following this one.  If we skipped creating
> .ready segments due to a crash, we rely on RemoveOldXlogFiles() to
> create them as needed in the end-of-recovery checkpoint.  If a record
> fits perfectly in the end of a segment, we'll still register it as a
> boundary for the next segment (hence why we use XLByteToSeg() instead
> of XLByteToPrevSeg()).  If database activity stops completely, there
> shouldn't be anything to mark ready.

The only way .ready files are created is that XLogNotifyWrite() is
called.  For regular WAL files during regular operation, that only
happens in XLogNotifyWriteSeg().  That, in turn, only happens in
NotifySegmentsReadyForArchive().  But if the system runs and never
writes WAL records that cross WAL boundaries, that function will see
that both earliestSegBoundary and latestSegBoundary are MaxXLogSegno,
and return without doing anything.  So no segments will be notified.

-- 
Álvaro Herrera           39°49'30"S 73°17'W  —  https://www.EnterpriseDB.com/