From: Tom Lane <tgl@sss.pgh.pa.us>
To: Magnus Hagander <magnus@hagander.net>
cc: Andres Freund <andres@anarazel.de>,
        Michael Banck <michael.banck@credativ.de>,
        David Steele <david@pgmasters.net>,
        PostgreSQL-development <pgsql-hackers@postgresql.org>,
        pgsql-committers@lists.postgresql.org
Subject: Re: pgsql: Validate page level checksums in base backups
In-reply-to: 
 <CABUevEz0uBqg_uyfA-yFiL6Wo=kjn5kE9tEQrJ-vwWug9Vrwfw@mail.gmail.com>
References: <E1f3KUE-0003rM-6g@gemulon.postgresql.org>
 <20180403171356.v4ji5sfhbc4sggij@alap3.anarazel.de>
 <CABUevExQa00NOeZd7N7faSZOnL+gtThKSDv2d5zxAP7Cp9YtLA@mail.gmail.com>
 <22322.1522780151@sss.pgh.pa.us>
 <CABUevEz0uBqg_uyfA-yFiL6Wo=kjn5kE9tEQrJ-vwWug9Vrwfw@mail.gmail.com>
Comments: In-reply-to Magnus Hagander <magnus@hagander.net>
	message dated "Tue, 03 Apr 2018 20:48:08 +0200"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <23573.1522781707.1@sss.pgh.pa.us>
Date: Tue, 03 Apr 2018 14:55:07 -0400
Message-ID: <23574.1522781707@sss.pgh.pa.us>
Precedence: bulk

Magnus Hagander <magnus@hagander.net> writes:
> On Tue, Apr 3, 2018 at 8:29 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> It's scribbling on the source cluster's disk files and assuming that that
>> translates one-for-one to what gets sent to the slave server --- but what
>> if some of the blocks that it modifies on-disk are resident in the
>> source's shared buffers?  I think you'd have to shut down the source and
>> then apply the corruption if you want stable results.

> It doesn't actually use a slave server as part of the tests.
> And basebackups don't read from the sources shared buffers, but it *does*
> read from the kernel buffers.

Right, so the failure case occurs when the source server writes back a
dirty page from its shared buffers to disk, in between the test script's
corrupting that page on-disk and then trying to read it.

Again, you'd have several orders of magnitude better chance of getting
reproducible behavior if you were targeting something that wasn't a
system catalog.

			regards, tom lane