Stanislaw:
On Thu, 28 Jul 2022 at 20:57, PG Bug reporting form
<noreply@postgresql.org> wrote:
> create table ss_temp_table1 as select id, v1, v2 from ss_d order by id desc
> limit 100000;
...
> FROM ss_d AS d
> JOIN ss_temp_table1 AS t1 ON t1.id = d.id AND t1.v1 = d.v1 AND t1.v2
> = d.v2
Do your real tables/queries have the same kind of redudant condition (
i.e., "AND t1.v1 = d.v1 AND t1.v2" is redundant due to id being a "PK"
)?
If so you could try omitting it ( and maybe try w & w/o seqscan
enabled too ) to see if this condition is leading the optimizer to
believe a plan is much cheaper than it really is.
If not you may be hitting a different problem with similar appearance.
Francisco Olarte.