Re: Remaining case where reltuples can become distorted across multiple VACUUM operations

On Mon, Aug 8, 2022 at 8:33 AM Matthias van de Meent
<boekewurm+postgres(at)gmail(dot)com> wrote:
> For example, if currently the measured 2% of the pages contains more
> than 100% of the previous count of tuples, or with your patch the last
> page contains more than 100% of the previous count of the tuples, that
> new count is ignored, which seems silly considering that the vacuum
> count is supposed to be authorative.

The 2% thing is conditioned on the new relpages value precisely
matching the existing relpages from pg_class -- which makes it very
targeted. I don't see why scanned_tuples greatly exceeding the
existing reltuples from pg_class is interesting (any more interesting
than the other way around).

We'll always accept scanned_tuples as authoritative when VACUUM
actually scans all pages, no matter what. Currently it isn't possible
for VACUUM to skip pages in a table that is 32 pages or less in size.
So even the new "single page" thing from the patch cannot matter
there.

--
Peter Geoghegan