On 2022-Jul-25, Pierre wrote:
> On Monday, July 25, 2022 11:52:41 AM CEST Alvaro Herrera wrote:
> > It seems that this would be too noisy to be truly usable. What if we
> > emitted a log line when the variable changed, and the value that's in
> > use when the connection starts?
>
> Then the log files would be filled by these messages, only to be able to make
> use of the few slow queries that are logged during the day.
Ah, yeah, that's not useful for that case ...
> Or it would need to be a log before each slow query. I'm not sure how
> well it would work.
... and this would probably be prohibitively complex to implement and
use, as well as too slow for the high traffic case.
Maybe your idea of allowing arbitrary GUCs is not a bad one, something
like
%{search_path}G
(where we add a letter at the end just so we can add other things in the
future that aren't GUCs.)
--
Álvaro Herrera PostgreSQL Developer — https://www.EnterpriseDB.com/