Re: Test slots invalidations in 035_standby_logical_decoding.pl only if dead rows are removed

Hi Alexander,

On Sat, Jun 08, 2024 at 07:00:00AM +0300, Alexander Lakhin wrote:
> Hello Bertrand and Michael,
>
> 23.01.2024 11:07, Bertrand Drouvot wrote:
> > On Tue, Jan 23, 2024 at 02:50:06PM +0900, Michael Paquier wrote:
> >
> > > Anyway, that's not the end of it. What should we do for snapshot
> > > snapshot records coming from the bgwriter?
> > What about?
> >
> > 3) depending on how stabilized this test (and others that suffer from "random"
> > xl_running_xacts) is, then think about the bgwriter.
>
> A recent buildfarm failure [1] reminds me of that remaining question.
>
> It's hard to determine from this info, why row_removal_activeslot was not
> invalidated, but running this test on a slowed down Windows VM, I (still)
> get the same looking failures caused by RUNNING_XACTS appeared just before
> `invalidating obsolete replication slot "row_removal_inactiveslot"`.
> So I would consider this failure as yet another result of bgwriter activity
> and add it to the list of known failures as such...

Thanks for the report! I think it makes sense to add it to the list of known
failures.

One way to deal with those corner cases could be to make use of injection points
around places where RUNNING_XACTS is emitted, thoughts?

Regards,

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