Re: Failure of subscription tests with topminnow

Ajin Cherian <itsajin@gmail.com>

From: Ajin Cherian <itsajin@gmail.com>
To: Amit Kapila <amit.kapila16@gmail.com>
Cc: Michael Paquier <michael@paquier.xyz>, Postgres hackers <pgsql-hackers@lists.postgresql.org>
Date: 2021-08-24T12:58:35Z
Lists: pgsql-hackers
On Tue, Aug 24, 2021 at 9:08 PM Amit Kapila <amit.kapila16@gmail.com> wrote:

> What happens when there is neither a new walsender nor an old
> walsender is present? It means to run the above statement when a new
> walsender is exited due to error "... slot is active ..." and before a
> new walsender could start. Also, allow old walsender (due to which the
> error occurs) to exit before executing the statement.
>
I tried this, then the query returns a null instead of 'false' because
there is no entry for that application_name.

postgres=# select pid != 31876 from pg_stat_replication where
application_name = 'sub';
 ?column?
----------
(0 rows)


> Also, it seems this failure happens on REL_11_STABLE branch, not sure
> if that matters, but it is better to use the same branch if you are
> using a different branch to reproduce the issue.
>

Ok, I didn't realise this. I will try this.

regards,
Ajin Cherian
Fujitsu Australia



Commits

  1. Fix the random test failure in 001_rep_changes.