David Garamond <lists(at)zara(dot)6(dot)isreserved(dot)com> writes:
> The table contain +- 1 mil records, all of the actual version of the
> queries below return < 10 rows, so an index should be used. Using an
> index scan, the query ran < 100ms. Using seq scan, 2-3 secs. So there is
> no good reason why a seq scan should be used, especially in a case of
> b='foo' or b='bar'.
[shrug...] We can't possibly diagnose a bad-plan-choice problem with
the amount of information you've provided. See
http://techdocs.postgresql.org/guides/SlowQueryPostingGuidelines
regards, tom lane