Message-ID: <534f1b4c-ad0c-6ba0-ed9a-521f5e240e0c@inbox.ru>
Date: Tue, 14 Feb 2023 18:52:20 +0300
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.7.1
Subject: Re: odd buildfarm failure - "pg_ctl: control file appears to be
 corrupt"
From: "Anton A. Melnikov" <aamelnikov@inbox.ru>
To: Thomas Munro <thomas.munro@gmail.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Andres Freund <andres@anarazel.de>,
 pgsql-hackers@postgresql.org
References: <20221123014224.xisi44byq3cf5psi@awork3.anarazel.de>
 <CA+hUKGL9w0VhZ=ceMDjXgCXy1BC5VzjwXc-HetSxfNzvBLge3A@mail.gmail.com>
 <CA+hUKGKr0TBwFMZ_BxcOps4_WDxhX5tS2D1Mwb1fAHyMKxo-UA@mail.gmail.com>
 <3748783.1669241103@sss.pgh.pa.us>
 <CA+hUKGK-BEe38aKNqHJDQ86LUW-CMwF5F9bo1JtJVg71FoDv_w@mail.gmail.com>
 <20d45197-80be-4984-33a9-a8d13708992a@inbox.ru>
 <CA+hUKGKzODWydEc0ZyAYXyFscqSQAvxDWCkDEOggiP4OZdr9BQ@mail.gmail.com>
 <CA+hUKGJBrKENOmq6zHGMOh637cjHL6VeOc1NF98Vtq9c962oBA@mail.gmail.com>
 <2a64f915-213b-4bd0-1cdf-bd987694021e@inbox.ru>
 <CA+hUKGKZoCCn+6Z_dmwQsKu5PVLbDu6G7fD6JU1qok84r1ry0A@mail.gmail.com>
 <d1a8f3a8-3ef9-1cd7-f4e9-ab1712c14745@inbox.ru>
Content-Language: en-US
In-Reply-To: <d1a8f3a8-3ef9-1cd7-f4e9-ab1712c14745@inbox.ru>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Archived-At: <https://www.postgresql.org/message-id/534f1b4c-ad0c-6ba0-ed9a-521f5e240e0c%40inbox.ru>
Precedence: bulk

Hi, Thomas!

On 14.02.2023 06:38, Anton A. Melnikov wrote:
> Also i did several experiments with fsync=on and found more appropriate behavior:
> The stress test with sizeof(ControlFileData) = 512+8 = 520 bytes failed in a 4,5 hours,
> but the other one with ordinary sizeof(ControlFileData) = 296 not crashed in more than 12 hours.

Nonetheless it crashed after 18 hours:

2023-02-13 18:07:21.476 MSK [7640] LOG:  starting PostgreSQL 16devel, compiled by Visual C++ build 1929, 64-bit
2023-02-13 18:07:21.483 MSK [7640] LOG:  listening on IPv6 address "::1", port 5432
2023-02-13 18:07:21.483 MSK [7640] LOG:  listening on IPv4 address "127.0.0.1", port 5432
2023-02-13 18:07:21.556 MSK [1940] LOG:  database system was shut down at 2023-02-13 18:07:12 MSK
2023-02-13 18:07:21.590 MSK [7640] LOG:  database system is ready to accept connections
@@@@@@@@@@@@@ sizeof(ControlFileData) = 296
2023-02-13 18:12:21.545 MSK [9532] LOG:  checkpoint starting: time
2023-02-13 18:12:21.583 MSK [9532] LOG:  checkpoint complete: wrote 3 buffers (0.0%); 0 WAL file(s) added, 0 removed, 0 recycled; write=0.003 s, sync=0.009 s, total=0.038 s; sync files=2, longest=0.005 s, average=0.005 s; distance=0 kB, estimate=0 kB; lsn=0/17AC388, redo lsn=0/17AC350
2023-02-14 12:12:21.738 MSK [8676] ERROR:  calculated CRC checksum does not match value stored in file
2023-02-14 12:12:21.738 MSK [8676] CONTEXT:  SQL statement "SELECT pg_control_system()"
	PL/pgSQL function inline_code_block line 1 at PERFORM
2023-02-14 12:12:21.738 MSK [8676] STATEMENT:  do $$ begin loop perform pg_control_system(); end loop; end; $$;


So all of the following is incorrect:

> Seems in that case the behavior corresponds to msdn.  So if it is possible
> to use fsync() under windows when the GUC fsync is off it maybe a solution
> for this problem. If so there is no need to lock the pg_control file under windows at all.

and cannot be a solution.


Sincerely yours,

-- 
Anton A. Melnikov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company