Horton

Re: Show WAL write and fsync stats in pg_stat_io

Michael Paquier <michael@paquier.xyz>

View thread
From: Michael Paquier <michael@paquier.xyz>
To: Nazir Bilal Yavuz <byavuz81@gmail.com>
Cc: Andres Freund <andres@anarazel.de>, pgsql-hackers <pgsql-hackers@postgresql.org>, "bharath.rupireddyforpostgres@gmail.com" <bharath.rupireddyforpostgres@gmail.com>
Date: 2023-11-20T07:47:19Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Fix copy-paste error related to the autovacuum launcher in pgstat_io.c

  2. Move SQL tests of pg_stat_io for WAL data to recovery test 029_stats_restart

  3. Add data for WAL in pg_stat_io and backend statistics

  4. Improve comment on top of pgstat_count_io_op_time()

  5. Refactor pgstat_prepare_io_time() with an input argument instead of a GUC

On Thu, Nov 09, 2023 at 02:39:26PM +0300, Nazir Bilal Yavuz wrote:
> There are some differences between pg_stat_wal and pg_stat_io while
> collecting WAL stats. For example in the XLogWrite() function in the
> xlog.c file, pg_stat_wal counts wal_writes as write system calls. This
> is not something we want for pg_stat_io since pg_stat_io counts the
> number of blocks rather than the system calls, so instead incremented
> pg_stat_io by npages.
> 
> Could that cause a problem since pg_stat_wal's behaviour will be
> changed? Of course, as an alternative we could change pg_stat_io's
> behaviour but in the end either pg_stat_wal's or pg_stat_io's
> behaviour will be changed.

Yep, that could be confusing for existing applications that track the
information of pg_stat_wal.  The number of writes is not something
that can be correctly shared between both.  The timings for the writes
and the syncs could be shared at least, right?

This slightly relates to pgstat_count_io_op_n() in your latest patch,
where it feels a bit weird to see an update of
PendingWalStats.wal_sync sit in the middle of a routine dedicated to
pg_stat_io..  I am not completely sure what's the right balance here,
but I would try to implement things so as pg_stat_io paths does not
need to know about PendingWalStats.
--
Michael