Re: pg_recvlogical: Prevent flushed data from being re-sent after restarting replication

Mircea Cadariu <cadariu.mircea@gmail.com>

View thread

From: Mircea Cadariu <cadariu.mircea@gmail.com>

To: Fujii Masao <masao.fujii@gmail.com>

Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>

Date: 2025-11-24T09:03:41Z

Lists: pgsql-hackers

Attachments

v1-0001-Add-test-for-pg_recvlogical-reconnection.patch (text/plain)

Hi,


An update: I have two topics from the review perspective.

On the test I proposed to be added to 030_pg_recvlogical.pl, I found a 
way to write it without using sleeps (which risk flakyness in CI). I've 
attached it as a patch for your consideration. I checked the test in the 
following way: on master it fails, but with your patch it passes.

Secondly I noticed in function sendFeedback at line 166, the startpos is 
set to output_written_lsn. This seems to counter conceptually the change 
you made in the patch, however it seems to not affect correctness. Shall 
we remove this line as to avoid confusion?