Re: Logical Replication not working for few Tables - Mailing list pgsql-bugs

From Amit Kapila
Subject Re: Logical Replication not working for few Tables
Date
Msg-id CAA4eK1Je2jBF3rwN3Qd8sdVoC0pJSDKe8k2HnZ1nYyk6JOJMww@mail.gmail.com
Whole thread Raw
In response to Logical Replication not working for few Tables  (Padmakumar Kadayaprth <padmakumark26@gmail.com>)
Responses Re: Logical Replication not working for few Tables
List pgsql-bugs
On Fri, Nov 12, 2021 at 8:06 PM Padmakumar Kadayaprth
<padmakumark26@gmail.com> wrote:
>
> Hi All,
>    I configured  logical replication some time back .Now except for 4 table replication is happening.Few weeks back
replicationgot stopped because of some conflict and these 4 tables were part of that conflict.  Someone resolved the
conflictbut replication has not started for these four tables .I dropped these tables from publication and added it
backand refreshed the subscription but still replication is not happening for these four . 
>

The refresh won't do anything if the tables have already been
replicated previously.

> For testing I created a new table and added it to publication and refreshed subscription , that was sussues.
>
> Is there any system table where these 4 tables are marked as conflict and not to be replicated?
>

No.

> Why is replication not happening only to those four tables.is this a bug?
>

It is quite surprising that replication changes for a few tables is
getting skipped whereas it is successful for others. One guess is that
new changes are updates and deletes of rows that didn't exist in the
tables on subscriber-side, as one might have removed/changed those
rows while resolving conflicts. If that happens then we log the
message at DEBUG1 log level.

I think you can try to construct an independent test on the same lines
and resolve conflict in a similar way and then see what is happening?


--
With Regards,
Amit Kapila.



pgsql-bugs by date:

Previous
From: Andres Freund
Date:
Subject: Re: BUG #17255: Server crashes in index_delete_sort_cmp() due to race condition with vacuum
Next
From: Masahiko Sawada
Date:
Subject: Re: References to parameters by name are lost in INSERT INTO ... SELECT .... statements in case of routines with the SQL-standard function body