Re: Parallel vacuum: workers together with leader could overrun cost budget

Hi,

On Wed, Oct 28, 2026 at 1:43 PM Masahiko Sawada <sawada(dot)mshk(at)gmail(dot)com> wrote:
>
> > I would just cap it at 1s like this:
>
> > + msec = Min(msec, Min(vacuum_cost_delay * 4 * nworkers, 1000));
>
> Sound reasonable change.

Thanks. Done that in the attached v3 patch.

> > I tend to agree with your point on not back-patching to <= PG18. As
> > mentioned upthread [1], the fact that it hasn't been reported so far
> > means it may have gone unnoticed, or everyone wanted vacuum to run
> > faster at the cost of foreground path performance.
>
> I'd like to withdraw my previous proposal to backpatch it to all
> supported versions. We haven't received any complaints about this
> behavior so far, and with the fix VACUUM could get slower after a
> minor update. So I agree to push it to v19 and master, and I think we
> should mention this behavior change in the v19 release notes.

Agreed.

> Also, I think it'd be better to update the paragraph in maintenance.sgml
> that says parallel workers use the same cost delay parameters as the
> leader, to mention that they share the cost limit.

Done.

Please find the attached v3 patch.

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