> The following query needs about 10 seconds to execute on a table with
> around 7 Mio. lines:
>
> select * from lin_albaran_venta where nro_factura = '0000001' order by
> nro_albaran, nro_linea;
> Plan:
> PLAN (LIN_ALBARAN_VENTA ORDER PK_LIN_ALBARAN_VENTAPK INDEX
> (FK_LIN_ALBARAN_VENTA_13))
>
> When adding field nro_factura to the order by, the query speeds up to 0,2
> secs:
>
> select * from lin_albaran_venta where nro_factura = '0000001' order by
> nro_factura, nro_albaran, nro_linea;
> Plan:
> PLAN SORT ((LIN_ALBARAN_VENTA INDEX (FK_LIN_ALBARAN_VENTA_13)))
>
> nro_factura is a FK, and (nro_albaran, nro_linea) the PK of table
> LIN_ALBARAN_VENTA.
>
> We recently upgraded from FB 2.1 to 2.5, and on 2.1 we didn't run into
> this problem.
>
> Is there any solution to this, perhaps in the configuration file, or do we
> have to revise
> all the order by's in our application, or perhaps even go back to FB 2.1?

Have you updated the statistics on the indices of your database. Try and see if that improves things.

Mark
--
Empfehlen Sie GMX DSL Ihren Freunden und Bekannten und wir
belohnen Sie mit bis zu 50,- Euro! https://freundschaftswerbung.gmx.de