Re: performance on update table from a join

Jean-Luc Lachance <jllachan@nsd.ca> writes:
> Hash Join  (cost=109.44..118460.53 rows=1 width=857)
>   ->  Seq Scan on c  (cost=0.00..112409.04 rows=1156604 width=519)
>   ->  Hash  (cost=109.44..109.44 rows=1 width=338)
>         ->  Merge Join  (cost=0.00..109.44 rows=1 width=338)
>               ->  Index Scan using routes_str_mun on routes r 
> (cost=0.00..52.00 rows=1000 width=152)
>               ->  Index Scan using rs_stname on rs s  (cost=0.00..52.00
> rows=1000 width=186)

Have you vacuum analyzed routes and rs?  I always disbelieve any plan
with 'rows=1000' in it, because I know that's the default assumption
when no stats are available...
        regards, tom lane