Re: START_REPLICATION SLOT causing a crash in an assert build

On Thu, Oct 13, 2022 at 1:21 AM Andres Freund <andres(at)anarazel(dot)de> wrote:
>
> Hi,
>
> On 2022-10-11 17:10:52 +0900, Masahiko Sawada wrote:
> > +# Reset the replication slot statistics.
> > +$node->safe_psql('postgres',
> > + "SELECT pg_stat_reset_replication_slot('regression_slot');");
> > +my $result = $node->safe_psql('postgres',
> > + "SELECT * FROM pg_stat_replication_slots WHERE slot_name = 'regrssion_slot'"
> > +);
>
> Typo in the slot name "regrssion_slot" instead of "regression_slot". We can't
> use * here, because that'll include the reset timestamp.

Fixed.

>
>
> > +# Teardown the node so the statistics is removed.
> > +$pg_recvlogical->kill_kill;
> > +$node->teardown_node;
> > +$node->start;
>
> ISTM that removing the file instead of shutting down the cluster with force
> would make it a more targeted test.

Agreed.

>
>
> > +# Check if the replication slot statistics have been removed.
> > +$result = $node->safe_psql('postgres',
> > + "SELECT * FROM pg_stat_replication_slots WHERE slot_name = 'regrssion_slot'"
> > +);
> > +is($result, "", "replication slot statistics are removed");
>
> Same typo as above. We can't assert a specific result here either, because
> recvlogical will have processed a bunch of changes. Perhaps we could check at
> least that the reset time is NULL?

Agreed.

>
>
> > +# Test if the replication slot staistics continue to be accumulated even after
>
> s/staistics/statistics/

Fixed.

I've attached an updated patch. I've added the common function to
start pg_recvlogical and wait for it to become active. Please review
it.

Regards,

--
Masahiko Sawada
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com