Csaba Nagy <nagy(at)ecircle-ag(dot)com> writes:
> Looking at Plan 2, it looks like the "limit" step is estimating wrongly
> it's cost.
The reason you get a bad plan is that this rowcount estimate is so far
off:
> -> Index Scan using pk_table_a on table_a ta (cost=0.00..324786.18 rows=388532 width=16) (actual time=454.389..460.138 rows=2 loops=1)
> Index Cond: (a = $1)
Raising the stats target helped no doubt because it didn't overestimate
the number of rows so much...
regards, tom lane