Re: 9.2.1 & index-only scans : abnormal heap fetches after VACUUM FULL

On 28 November 2013 22:17, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:

> The fact that you've needed to modify SetHintBits() to make this work
> is pretty nasty. I'm not exactly sure what to do about that, but it
> doesn't seem good.

That makes me feel bad also.

I think we should be looking for some special case routines rather
than piggybacking there.

Wonderful to see you personally and directly addressing this concern
and very supportive of this solution for 9.4.

I wonder about a user option to let these routines wait until they are
the oldest snapshot, so they can then set every page as all-visible
without even bothering to check individual items. Or maybe a user
option to set them all-visible even without actually being the oldest,
as is possible with COPY FREEZE. Full lock on the table is a big
thing, so a shame to waste the opportunity to set everything visible
just because some long running task is still executing somewhere (but
clearly not here otherwise we wouldn't have the lock).

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