Hi,
On 2025-03-03 17:44:54 +0100, Álvaro Herrera wrote:
> On 2025-Mar-03, Melanie Plageman wrote:
>
> > Add relallfrozen to pg_class
> >
> > Add relallfrozen, an estimate of the number of pages marked all-frozen
> > in the visibility map.
> >
> > pg_class already has relallvisible, an estimate of the number of pages
> > in the relation marked all-visible in the visibility map. This is used
> > primarily for planning.
> >
> > relallfrozen, together with relallvisible, is useful for estimating the
> > outstanding number of all-visible but not all-frozen pages in the
> > relation for the purposes of scheduling manual VACUUMs and tuning vacuum
> > freeze parameters.
>
> I'm confused about this. Why was the new value added to pg_class
> instead of to the pgstat system? I don't think relallvisible is a good
> precedent, because as you write here, that one is used for planning,
> which has different requirements. For vacuuming metrics we rely on
> pgstat.
We do so because updating the other stats would cause too much churn - the
number of dead tuples etc changes rather rapidly. That's not the case
here. Having reliable information after a crash (where we loose pgstats) that
95% of the table is actually frozen, rather than not seems good.
Greetings,
Andres Freund
