On Fri, Mar 8, 2024, at 6:44 AM, Hayato Kuroda (Fujitsu) wrote: > Hmm, right. I considered below improvements. Tomas and Euler, how do you > think?
I'm posting a new patchset v28. I changed the way that the check function works. From the usability perspective, it is better to test all conditions and reports all errors (if any) at once. It avoids multiple executions in dry run mode just to figure out all of the issues in the initial phase. I also included tests for it using Shlok's idea [1] although I didn't use v27-0004. Shlok [1] reported that it was failing on Windows since the socket-directory option was added. I added a fix for it. Tomas pointed out the documentation [2] does not provide a good high level explanation about pg_createsubscriber. I expanded the Description section and moved the prerequisites to Nodes section. The prerequisites were grouped into target and source conditions on their own paragraph instead of using a list. It seems more in line with the style of some applications. As I said in a previous email [3], I removed the retain option. [1] https://www.postgresql.org/message-id/canhcyeu4q3dwh9ax9bpojcm4ebbhyfenegoaz8xfgyjmcpz...@mail.gmail.com [2] https://www.postgresql.org/message-id/6423dfeb-a729-45d3-b71e-7bf1b3adb0c9%40enterprisedb.com [3] https://www.postgresql.org/message-id/e390e35e-508e-4eb8-92e4-e6b066407a41%40app.fastmail.com -- Euler Taveira EDB https://www.enterprisedb.com/
v28-0001-pg_createsubscriber-creates-a-new-logical-replic.patch.gz
Description: application/gzip
v28-0002-Use-last-replication-slot-position-as-replicatio.patch.gz
Description: application/gzip
v28-0003-port-replace-int-with-string.patch.gz
Description: application/gzip