Horton

Re: Introduce a new view for checkpointer related stats

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Andres Freund <andres@anarazel.de>
Cc: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>, "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>, Robert Haas <robertmhaas@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2023-02-10T05:59:13Z
Lists: pgsql-hackers 
On Thu, Feb 09, 2023 at 04:46:04PM -0800, Andres Freund wrote:
> I think if we end up breaking compat, we should just drop that
> column.

Indeed.

> Yep, and I think you are all wrong, and that this is just going to cause
> unnecessary pain :). I'm not going to try to prevent the patch from going in
> because of this, just to be clear.

Catalog attributes have faced a lot of renames across the years, with
the same potential of breakages for monitoring tools.  I am not saying
that all of them are justified, but we have usually done so because it
makes sense to reshape things in the way they are now, thinking
long-term.  Splitting pg_stat_bgwriter into two views does not strike
me as something that bad, TBH, because it becomes clearer which stats
are attached to which process (bgwriter or checkpointer).  (Note: I
have not checked in details the stats switching to the new view and
how pertinent each choice is.)
--
Michael

Commits

  1. Expand regression tests for pg_stat_reset_shared()

  2. Introduce pg_stat_checkpointer

  3. Remove buffers_backend and buffers_backend_fsync from pg_stat_checkpointer