Hi,
On Mon, Mar 21, 2022 at 02:44:08AM +0000, PG Bug reporting form wrote:
> For the select command to use the brin index of the id2 field, I have to
> change enable_seqscan = off
>
> create temp table teste as
> select s::int as id1, s::int as id2, quote_ident('Nome ' || to_char(s,
> 'FM00000000'))::varchar(15) as nome
> from generate_series(1,10000000) x(s);
>
> create index teste1 on teste (id1);
> create index teste2 on teste using brin (id2);
>
> EXPLAIN (analyze on, wal on, verbose on, format text)
> select id1 from teste where id1 = 1000000;
>
> EXPLAIN (analyze on, wal on, verbose on, format text)
> select id2 from teste where id2 = 1000000;
This is not a bug. You can't expect postgres to chose a good plan if you don't
have any statistics. You should do an explicit "ANALYZE teste" after loading
your data (same for any bulk loading operation).