Horton

Re: Add support for specifying tables in pg_createsubscriber.

Euler Taveira <euler@eulerto.com>

View thread
From: "Euler Taveira" <euler@eulerto.com>
To: "Peter Smith" <smithpb2250@gmail.com>, "vignesh C" <vignesh21@gmail.com>
Cc: tianbing <tian_bing_0531@163.com>, "'Shubham Khanna'" <khannashubham1197@gmail.com>, "Chao Li" <li.evan.chao@gmail.com>, "houzj.fnst@fujitsu.com" <houzj.fnst@fujitsu.com>, "kuroda.hayato@fujitsu.com" <kuroda.hayato@fujitsu.com>, "PostgreSQL Hackers" <pgsql-hackers@lists.postgresql.org>
Date: 2025-12-09T23:31:56Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Support existing publications in pg_createsubscriber.

Attachments

On Sun, Dec 7, 2025, at 11:52 PM, Peter Smith wrote:
>
> OK. Done as suggested in v23.
>

I took another look at this patch. I noticed that it increased the test time in
3 or 4s. I started writing the suggestions to this email but since it increases
a lot, it is better to provide a patch and explain the main points. Regarding
the tests, there are 2 new nodes (one used to test this feature -- node_s2 --
and another one with no real function -- node_s3). There is no need to add new
nodes, just use the created publication into one of the existing tests. That's
what I did (see test_pub3). I also added a new table to ensure the publication
only contains tbl1 and not not_replicated table. I removed the dry run part
because it only increases the total time and doesn't improve coverage. Finally,
coverage stays the same and a small increment in the execution time (less than
1s).

$ tail -n 8 src/bin/pg_basebackup/pg_createsubscriber.c.gcov.out
File 'pg_createsubscriber.c'
Lines executed:83.81% of 951
Branches executed:93.15% of 467
Taken at least once:74.73% of 467
Calls executed:73.48% of 675
Creating 'pg_createsubscriber.c.gcov'

Lines executed:83.81% of 951

I also changed the code a bit. In particular, I don't like
check_publication_exists and find_publication seems cleaner to me so I renamed
it. I replaced the psprintf because the pattern is to use PQExpBuffer functions
for queries. Added pg_catalog to the query. I changed pg_fatal with pg_log_error
plus disconnect_database -- see previous discussion about it. I added a comment
explaining why made_publication = false. Finally, I refactored the logic that
drops publications; the if (!drop_all_pubs || dry_run) is rather confusing.
Documentation was also changed a bit. One paragraph instead of two because it is
talking about the same topic.

Here is your v23 plus the fixes that I described above. IMO it is good to go so
I set it to Ready for Committer. Additional comments are always welcome.


-- 
Euler Taveira
EDB   https://www.enterprisedb.com/