Horton

Re: Corruption during WAL replay

Robert Haas <robertmhaas@gmail.com>

View thread View on postgr.es
From: Robert Haas <robertmhaas@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Kyotaro Horiguchi <horikyota.ntt@gmail.com>, Ibrar Ahmed <ibrar.ahmad@gmail.com>, tejeswarm@hotmail.com, Andres Freund <andres@anarazel.de>, hlinnaka <hlinnaka@iki.fi>, Masahiko Sawada <masahiko.sawada@2ndquadrant.com>, "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>, Daniel Wood <hexexpert@comcast.net>
Date: 2021-09-24T20:08:44Z
Lists: pgsql-hackers 
On Fri, Sep 24, 2021 at 3:42 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
> > I like this patch.
>
> I think the basic idea is about right, but I'm not happy with the
> three-way delayChkpt business; that seems too cute by three-quarters.
> I think two independent boolean flags, one saying "I'm preventing
> checkpoint start" and one saying "I'm preventing checkpoint completion",
> would be much less confusing and also more future-proof.  Who's to say
> that we won't ever need both states to be set in the same process?

Nobody, but the version of the patch that I was looking at uses a
separate bit for each one:

+/* symbols for PGPROC.delayChkpt */
+#define DELAY_CHKPT_START (1<<0)
+#define DELAY_CHKPT_COMPLETE (1<<1)

One could instead use separate Booleans, but there doesn't seem to be
anything three-way about this?

-- 
Robert Haas
EDB: http://www.enterprisedb.com

Commits

  1. Harden TAP tests that intentionally corrupt page checksums.

  2. Fix possible recovery trouble if TRUNCATE overlaps a checkpoint.

  3. Remember to reset yy_start state when firing up repl_scanner.l.