Le 03/09/2019 à 15:43, Tom Lane a écrit :
> "Arnaud L." <arnaud.listes@codata.eu> writes:
>> We have upgraded our database from 9.6 to 11 (and updated PostGIS from
>> 2.3 to 2.5 as well).
>> ...
>
> Have you re-ANALYZEd the database? The problem with this query
> seems to be the spectacularly awful rowcount estimate here:
>
>> -> Bitmap Index Scan on planet_osm_ways_nodes_idx
>> (cost=0.00..11190.36 rows=1420982 width=0) (actual time=0.268..0.268
>> rows=1 loops=1)
>> Index Cond: (nodes && '{1}'::bigint[])
>
> The planner should be able to do better than that, given up-to-date
> statistics on the "nodes" column.
Tom, I can confirm that with up to date statistics the planner is still
lost.
I did a REINDEX to rule out a broken index and the estimate is still in
the 100k+ range.
Regards
--
Arnaud