Horton

Re: Enhance 'pg_createsubscriber' to retrieve databases automatically when no database is provided.

Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>

View thread View on postgr.es
From: Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>
To: vignesh C <vignesh21@gmail.com>
Cc: Shubham Khanna <khannashubham1197@gmail.com>, Nisha Moond <nisha.moond412@gmail.com>, Ajin Cherian <itsajin@gmail.com>, "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>, Peter Smith <smithpb2250@gmail.com>, Amit Kapila <amit.kapila16@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2025-03-20T11:29:00Z
Lists: pgsql-hackers 
On Thu, Mar 20, 2025 at 3:41 PM vignesh C <vignesh21@gmail.com> wrote:
>
> On Thu, 20 Mar 2025 at 10:25, Shubham Khanna
> <khannashubham1197@gmail.com> wrote:
> >
> >
> > I have created two patches, v16-0001 and v16-0002, to address the
> > performance issue. I conducted performance testing, and here are the
> > results:
> > - The difference in execution time between HEAD and the v15 patch was 53.2%.
> > - After removing the suggested test cases, the difference reduced to
> > 36.43%, showing a significant improvement.
> >
>
> It is still taking quite a while, can we compare with the following
> too see how much extra time it takes:
> a) remove insert and select verification for the all database
> successful tests as all of these are logical replication verification
> which is extensively tested

If the intent is to remove those, I don't think it's a good idea.
Other tests may be testing replication but they don't test the setup
created by --all, which consists of objects generated with names
crafted by utility.

Said that we should also test the behaviour when such objects already
exist. But I guess randomness in generated names is hard to control.

-- 
Best Wishes,
Ashutosh Bapat

Commits

  1. Cosmetic fixes for pg_createsubscriber's -all option.

  2. pg_createsubscriber: Add '--all' option.