On Tue, Feb 27, 2024 at 11:59:00PM +0500, Kirill Reshke wrote:
> Do we need to test the pg_cancel_backend vs autovacuum case at all?
> I think we do. Would it be better to split work into 2 patches: first one
> with tests against current logic, and second
> one with some changes/enhancements which allows to cancel running autovac
> to non-superuser (via `pg_signal_autovacuum` role or some other mechanism)?
If we need to add tests for pg_signal_backend, I think it's reasonable to
keep those in a separate patch from pg_signal_autovacuum.
--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com