Tino Schwarze <postgresql@tisc.de> wrote:
> I've seen PostgreSQL perform a lot worse after setting
> effective_cache_size to 2 GB on a 8 GB dedicated database system.
> The planner started ignoring indices and doing sequential scans.
> Lowering effective_cache_size to 512 MB solved that.
You'd probably get overall better plans by fixing that with other
configuration options. If the active portion of your database is
pretty well cached, you might try setting both random_page_cost and
seq_page_cost to 0.1. If your data isn't that well cached, you could
try leaving seq_page_cost at 1 but lower random_page_cost to somewhere
in the neighborhood of 2.
-Kevin