Le 05/04/2013 16:17, Tom Lane a écrit :
> Franck Routier <franck.routier@axege.com> writes:
>> Le 04/04/2013 21:08, Tom Lane a écrit :
>>> Hmph. Can't see any reason for that to take a remarkably long time to
>>> plan. Can you put together a self-contained test case demonstrating
>>> excessive planning time? What PG version is this, anyway?
>> What I notice is that removing joins has a huge impact on the time
>> explain takes to return:
> Hm, kind of looks like it's just taking an unreasonable amount of time
> to process each join clause. What have you got the statistics targets
> set to in this database? What are the datatypes of the join columns?
> And (again) what PG version is this exactly?
>
> regards, tom lane
>
>
Postgresql version is 8.4.8.
Value for default_statistics_target is 5000 (maybe this is the problem ?)
Join columns datatypes are all varchar(32).
Also notice that not all joins have equals imppact on the time taken:
removing join on some tables has no impact (sanact, rsaidp), while
removing joins on others (saneds, sandia) has an important effect...
Regards,
Franck