On Thu, Oct 12, 2023 at 02:00:00PM +0300, Alexander Lakhin wrote:
> So to fail on the test, skink should perform at least twice slower than
> usual, and may be it's an extraordinary condition indeed, but on the other
> hand, may be increase checkpoint_timeout as already done in several tests
> (015_promotion_pages, 038_save_logical_slots_shutdown, 039_end_of_wal, ...).
>
> [1] https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=skink&dt=2023-10-10%2017%3A10%3A11
> [2] https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=skink&dt=2022-11-07%2020%3A27%3A11
Thanks for the investigation. Increasing the checkpoint timeout is
not a perfect science but at least it would work until a machine is
able to be slower than the current limit reached, so I would be OK
with your suggestion and raise the bar a bit more to prevent the race
created by these extra checkpoints triggered because of the time.
--
Michael