Julien Cigar <jcigar(at)ulb(dot)ac(dot)be> wrote:
> I tried to play on the various cost settings but it's doesn't
> change anything, except setting random_page_cost to 1 (which will
> lead to bad plans for other queries, so not a solution)
Yeah, you clearly don't have the active portion of your database
fully cached, so you don't want random_page_cost to go as low as
seq_page_cost.
Here's one suggestion to try:
random_page_cost = 2
cpu_tuple_cost = 0.05
I have found that combination to work well for me when the level of
caching is about where you're seeing it. I am becoming increasingly
of the opinion that the default for cpu_tuple_cost should be higher
than 0.01.
Please let us know whether that helps.
-Kevin