Feedback-ID: id4a34324:Fastmail
Date: Fri, 24 Mar 2023 20:00:44 -0700
From: Andres Freund <andres@anarazel.de>
To: Michael Paquier <michael@paquier.xyz>
Cc: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>,
	Kyotaro Horiguchi <horikyota.ntt@gmail.com>,
	pgsql-hackers@lists.postgresql.org
Subject: Re: Reconcile stats in find_tabstat_entry() and get rid of
 PgStat_BackendFunctionEntry
Message-ID: <20230325030044.nhfas3nommsmwl6q@awork3.anarazel.de>
References: <9262b2a9-6785-1f83-bf7f-7fcf90c2c113@gmail.com>
 <20230215.095644.822493184688844040.horikyota.ntt@gmail.com>
 <130095b4-7c6d-b6f1-69d4-94f66231f81d@gmail.com>
 <20230216212121.uvav7xgdh2ysqjy5@awork3.anarazel.de>
 <f08fee25-1dac-b68b-19af-f4b55ceb3802@gmail.com>
 <ZBK3S5Na71pjgGCX@paquier.xyz>
 <621ed53c-4d9a-8f77-ab8c-20235f2d95a1@gmail.com>
 <ZBMBh5ve3tyqH5/m@paquier.xyz>
 <f6ea424b-fabe-b702-e4e3-b71ebc827d59@gmail.com>
 <ZB5i1lt3v/AUSvpq@paquier.xyz>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <ZB5i1lt3v/AUSvpq@paquier.xyz>
Archived-At: <https://www.postgresql.org/message-id/20230325030044.nhfas3nommsmwl6q%40awork3.anarazel.de>
Precedence: bulk

Hi,

On 2023-03-25 11:56:22 +0900, Michael Paquier wrote:
> On Thu, Mar 16, 2023 at 02:13:45PM +0100, Drouvot, Bertrand wrote:
> > On 3/16/23 12:46 PM, Michael Paquier wrote:
> >>> I don't think we should pay any particular attention to those 2 ones
> >>> as anyway nothing prevent the 7 others to be called outside of the
> >>> pg_stat_xact_all_tables view.
> >> 
> >> I am not quite sure, TBH.  Did you look at the difference with a long
> >> chain of subtrans, like savepoints?  The ODBC driver "loves" producing
> >> a lot of savepoints, for example.
> > 
> > No, I did not measure the impact.
> 
> I have been thinking again about this particular point, and I would be
> fine with an additional boolean flag to compute the subtrans data in
> the global counter when only necessary.  This would not make the
> macros for the stat functions that much more complicated, while being
> sure that we don't do unnecessary computations when we know that we
> don't need them..

I don't understand what we're optimizing for here. These functions are very
very very far from being a hot path. The xact functions are barely ever
used. Compared to the cost of query evaluation the cost of iterating throught
he subxacts is neglegible.

Greetings,

Andres Freund