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

Hi,

On Tue, Jan 23, 2024 at 02:50:06PM +0900, Michael Paquier wrote:
> On Mon, Jan 22, 2024 at 09:07:45AM +0000, Bertrand Drouvot wrote:
>
> I've rewritten some of that, and applied the patch down to v16.

Thanks!

> > Yeah, I can clearly see how this patch helps from a theoritical perspective but
> > rely on Alexander's testing to see how it actually stabilizes the test.
>
> Anyway, that's not the end of it. What should we do for snapshot
> snapshot records coming from the bgwriter?

I've mixed feeling about it. On one hand we'll decrease even more the risk of
seeing a xl_running_xacts in the middle of a test, but on the other hand I agree
with Tom's concern [1]: I think that we could miss "corner cases/bug" detection
(like the one reported in [2]).

What about?

1) Apply "wait_until_vacuum_can_remove() + autovacuum disabled" where it makes
sense and for tests that suffers from random xl_running_xacts. I can look at
031_recovery_conflict.pl, do you have others in mind?
2) fix [2]
3) depending on how stabilized this test (and others that suffer from "random"
xl_running_xacts) is, then think about the bgwriter.

[1]: https://www.postgresql.org/message-id/1375923.1705291719%40sss.pgh.pa.us
[2]: https://www.postgresql.org/message-id/flat/ZaTjW2Xh+TQUCOH0(at)ip-10-97-1-34(dot)eu-west-3(dot)compute(dot)internal

Regards,

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