Scott,
Le mercredi 7 Avril 2004 22:13, scott.marlowe a écrit :
> On Wed, 7 Apr 2004, [iso-8859-15] Hervé Piedvache wrote:
> >
> > I have a database with one table with about 22 millions records.
> > I have a script every day wich insert about 200 000 records ...
> > Sometime my script takes 4 hours to insert the data, sometime 1 hour.
>
> Are you wrapping all your inserts into one or several large transactions?
> This can make inserting data MUCH faster in postgresql.
It's one transaction ... because it's an INSERT from SELECT ...
> Do you have fk/pk relationships? If so, are they of the same types?
No no fk/pk ...
> I wouldn't worry about the vacuum / analyze numbers being a little off. A
> small variance of a few percent is no biggie, it's when things are off by
> factors that the planner gets confused and makes bad decisions.
Yes but he gets confused of about 150 000 rows ... it's few in front of the 22
millions ... but it's done just after an Analyze ... so why the analyze do
not get the information ??
Regards,
--
Hervé Piedvache
Elma Ingénierie Informatique
6 rue du Faubourg Saint-Honoré
F-75008 - Paris - France
Pho. 33-144949901
Fax. 33-144949902