Re: Synchronizing slots from primary to standby

Hi,

On Fri, Jan 12, 2024 at 03:46:00AM +0000, Zhijie Hou (Fujitsu) wrote:
> On Thursday, January 11, 2024 11:42 PM Bertrand Drouvot <bertranddrouvot(dot)pg(at)gmail(dot)com> wrote:
>
> Hi,
>
> > On Thu, Jan 11, 2024 at 04:22:56PM +0530, Amit Kapila wrote:
> > IIUC, this would be a sync slot (so not usable until promotion) that could not be
> > used anyway (invalidated), so I'll vote for drop / re-create then.
>
> Such race can happen when user drop and re-create the same failover slot on
> primary as well. For example, user dropped one failover slot and them
> immediately created a new one by copying from an old slot(using
> pg_copy_logical_replication_slot). Then the slotsync worker will find the
> restart_lsn of this slot go backwards.
>
> The steps:
> ----
> SELECT 'init' FROM pg_create_logical_replication_slot('isolation_slot', 'pgoutput', false, false, true);
> SELECT 'init' FROM pg_create_logical_replication_slot('test', 'pgoutput', false, false, true);
>
> - Advance the restart_lsn of 'test' slot
> CREATE TABLE test2(a int);
> INSERT INTO test2 SELECT generate_series(1,10000,1);
> SELECT slot_name FROM pg_replication_slot_advance('test', pg_current_wal_lsn());
>
> - re-create the test slot but based on the old isolation_slot.
> SELECT pg_drop_replication_slot('test');
> SELECT 'copy' FROM pg_copy_logical_replication_slot('isolation_slot', 'test');
>
> Then the restart_lsn of 'test' slot will go backwards.

Yeah, that's right.

BTW, I think it's worth to add those "corner cases" in the TAP tests related to
the sync slot feature (the more coverage the better).

Regards,

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