Re: Handle concurrent drop when doing whole database vacuum

Hi,

On Thu, Jun 25, 2026 at 3:23 PM Michael Paquier <michael(at)paquier(dot)xyz> wrote:
>
> On Thu, Jun 25, 2026 at 11:04:55AM -0700, Bharath Rupireddy wrote:
> > On Thu, Jun 25, 2026 at 10:41 AM surya poondla <suryapoondla4(at)gmail(dot)com> wrote:
> >>
> >> On the test front: once the race code path is gone, the injection point with the spec file isn't strictly needed either.
> >> A simpler test that shows "VACUUM with a concurrent drop successfully completes without error" should suffice.
> >> Bharath's TAP test suggestion under src/test/modules/test_misc seems like a good fit now.
> >
> > Thinking about this further, the test is NOT needed at all with this
> > approach because vacuum_open_relation() already protects against
> > concurrent drops. So, -1 for adding a new test.
>
> Well, as you mentioned upthread the only test infrastructure that we
> could use here is TAP, as database-wide VACUUMs could get costly when
> running other tests in parallel and we need stability. But I cannot
> get excited about the cost this would have versus the coverage we get.
> Cycles are better spent elsewhere, even if injection points could be
> leveraged here.

Agreed. I don't think it brings in any new coverage. Let's not
complicate things with injection points and tests for this fix.

--
Bharath Rupireddy
Amazon Web Services: https://aws.amazon.com