Re: AutoVacuum Behaviour Question

Bruce McAlister wrote:
> Alvaro Herrera wrote:
> > Bruce McAlister wrote:
> >
> > Ok now this is interesting:
> >
> >> select datname, age(datfrozenxid) from pg_database;
> >> datname | age
> >> -------------------------+-----------
> >> blueface-crm | 441746613
> >
> > Note this value is 440 million, and you said in your original report that
> >
> >> autovacuum_freeze_max_age = 200000000
> >
> > 200 million. So this database is being selected each time because of
> > this.
> >
>
> Ahhh okay, I didnt know how to extract the age for a database. Learnt
> something new here.
>
> > However, what should happen is that after the vacuum the age of the
> > database is decreased after the vacuuming. What's your
> > vacuum_freeze_min_age setting?
>
> My *_freeze_* values are:
>
> autovacuum_freeze_max_age = 200000000
> #vacuum_freeze_min_age = 100000000
>
> The vacuum_freeze_min_age is the default at 100 million (I assume).

What do you get from a SHOW vacuum_freeze_min_age? That would tell you
what's the actual value in use. Most likely it's those 100 million but
if you change it, reload, then comment it back in the file and reload
again, the value in use will be the one to which you first changed it.

> How much is the age decremented by on a vacuum run then?

It should be decremented to the vacuum_freeze_min_age. However, I'm
running some experiments with your settings and apparently it's not
working as it should.

--
Alvaro Herrera http://www.PlanetPostgreSQL.org/
"Hay que recordar que la existencia en el cosmos, y particularmente la
elaboración de civilizaciones dentro de él no son, por desgracia,
nada idílicas" (Ijon Tichy)