On Monday 20 May 2002 22:54, Tom Lane wrote:
> > listmembers | server_id | 0 | 4 | 1150 |
> > {34062,32715,42495,15182,38013,36503,13746,46532,42434,15392}
> >
> > {0.038,0.027,0.024,0.0143333,0.012,0.01,0.00933333,0.00766667,0.00733333,
> >0.007}
> >
> > | {12855,15419,16223,20598,26019,30733,34402,38139,40811,44986,49509}
> > | 0.428932
>
> Hmm. Correlation 0.43 is high enough to suggest that there's some
> clustering effect. If you look in the archives there's been prior
> discussion about whether to make the optimizer weight the correlation
> factor more strongly.
I have increased sequence scan weight 5 times. And it uses index scan now, but
this is not the solution. I will have a look in the archives.
--
Denis