Stephen Frost wrote:
> * Andres Freund (andres@2ndquadrant.com) wrote:
> > There's one argument for supporting more for VACUUM than the rest - it
> > can't be executed directly as the result of a query as the others
> > can... I wonder if that'd not better be answered by adding a feature to
> > vacuumdb that allows selecting the to-be-vacuumed table by a user
> > defined query.
>
> Wow. That's certainly an interesting idea.
+1.
> We might end up turning the autovacuum process into a generalized
> scheduler/cron-like entity that way though. I'd rather we just build
> that. Users would then be able to run a script periodically which
> would add VACUUM commands to be run on whichever tables they want to
> the jobs queue, either for immediate execution or at whatever time they
> want (or possibly chronically :).
This too. I think there's one or two orders of magnitude of difference
in implementation effort of these two ideas, however.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services