Re: Autovacuum fails to keep visibility map up-to-date in mostly-insert-only-tables

On 2014-10-19 12:50:30 -0400, Tom Lane wrote:
> Andres Freund <andres(at)2ndquadrant(dot)com> writes:
> > There's rub here though. We unconditionally do:
> > /* Do post-vacuum cleanup and statistics update for each index */
> > for (i = 0; i < nindexes; i++)
> > lazy_cleanup_index(Irel[i], indstats[i], vacrelstats);
>
> > and that's not particularly cheap. Maybe we should make that conditional
> > when there's been no lazy_vacuum_index/heap calls at all?
>
> Absolutely not. If the cleanup step is skippable, it would be the
> province of the index AM to make that decision.

Fair point. At the moment we're doing a full of nbtree indexes everytime
we do a vacuum. Even when the heap vacuum only scanned a couple hundred
pages of a huge table. That makes partial vacuum noticeably less
useful. So I do think we need to do something to improve upon the
situation.

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services