Re: undetected deadlock in ALTER SUBSCRIPTION ... REFRESH PUBLICATION

On Fri, Nov 24, 2023 at 5:05 PM Shlok Kyal <shlok(dot)kyal(dot)oss(at)gmail(dot)com> wrote:
>
> > I tried to reproduce the issue and was able to reproduce it with
> > scripts shared by Tomas.
> > I tried testing it from PG17 to PG 11. This issue is reproducible for
> > each version.
> >
> > Next I would try to test with the patch in the thread shared by Amit.
>
> I have created the v1 patch to resolve the issue. Have tested the
> patch on HEAD to PG12.
> The same patch applies to all the versions. The changes are similar to
> the one posted in the thread
> https://www.postgresql.org/message-id/1412708.1674417574%40sss.pgh.pa.us
>

(it's quite likely we hold lock on
+ * pg_replication_origin, which the sync worker will need
+ * to update).

This part of the comment is stale and doesn't hold true. You need to
update the reason based on the latest problem discovered in this
thread. I think you can compare the timing of regression tests in
subscription, with and without the patch to show there is no
regression. And probably some tests with a large number of tables for
sync with very little data.

--
With Regards,
Amit Kapila.