Re: compute_query_id and pg_stat_statements - Mailing list pgsql-hackers

From Julien Rouhaud
Subject Re: compute_query_id and pg_stat_statements
Date
Msg-id 20210427062504.ofz3rgedbrl5bep2@nol
Whole thread Raw
In response to Re: compute_query_id and pg_stat_statements  (Andres Freund <andres@anarazel.de>)
Responses Re: compute_query_id and pg_stat_statements  (Michael Paquier <michael@paquier.xyz>)
List pgsql-hackers
On Mon, Apr 26, 2021 at 11:37:45AM -0700, Andres Freund wrote:
> Hi,
> 
> On 2021-04-26 14:21:00 -0400, Tom Lane wrote:
> > Andres Freund <andres@anarazel.de> writes:
> > That's sounding like a pretty sane design, actually.  Not sure about
> > the shared-library-name-with-fixed-function-name detail, but certainly
> > it seems to be useful to separate "I need a query-id" from the details
> > of the ID calculation.
> > 
> > Rather than a GUC per se for the ID provider, maybe we could have a
> > function hook that defaults to pointing at the in-core computation,
> > and then a module wanting to override that just gets into the hook.
> 
> I have a preference to determining the provider via GUC instead of a
> hook because it is both easier to introspect and easier to configure.

In any case, having a different provider would greatly simplify third-party
queryid lib authors and users life.  For now the core queryid is computed
before post_parse_analyze_hook, but any third party plugin would have to do it
as a post_parse_analyze_hook, so you have to make sure that the lib is at the
right position in shared_preload_libraries to have it work, eg. [1], depending
on how pg_stat_statements and other similar module call
prev_post_parse_analyze_hook, which is a pretty bad thing.

> If the provider is loaded via a hook, and the shared library is loaded
> via shared_preload_libraries, one can't easily just turn that off in a
> single session, but needs to restart or explicitly load a different
> library (that can't already be loaded).

On the other hand we *don't* want to dynamically change the provider.
Temporarily enabling/disabling queryid calculation is ok, but generating
different have for the same query isn't.

> We also don't have any way to show what's hooking into a hook.

If we had a dedicated query_id hook, then plugins should error out if users
configured multiple plugins to calculate a query_id, so it should be easy to
know which plugin is responsible for it without knowing who hooked into the
hook.

[1] https://github.com/rjuju/pg_queryid/blob/master/pg_queryid.c#L116-L117



pgsql-hackers by date:

Previous
From: Michael Paquier
Date:
Subject: Small issues with CREATE TABLE COMPRESSION
Next
From: "tanghy.fnst@fujitsu.com"
Date:
Subject: RE: Truncate in synchronous logical replication failed