Emre Hasegeli wrote:
> >> From the above, may I presume that it is best to cluster (or sort), the
> >> table based on the intended
> >> BRIN column(s) before actually creating the index to insure the pages are
> >> adjacent? If so, should
> >> that not be included in the documentation, instead of implied?
>
> The same question is asked to me at PGConf.DE. I think it would be
> nice to address it in the documentation somehow. Maybe, we should
> also explain how the table is physically organised. It is not clear
> to users what kind of operations would make BRIN more useful.
Grumble.
> > I don't have faith in CLUSTER anyway. Taking exclusive locks and all.
>
> It also requires a btree index. If you can afford to have btree, you
> probably don't need BRIN anyway. Something lighter than CLUSTER which
> can use BRIN would be useful.
What I think would be useful is a way for the BRIN index to guide
location of a new tuple, so that it's put in the right spot right from
the start, instead of having it be moved later.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services