Patrick Hatcher <pathat@comcast.net> writes:
> Hash Join (cost=1246688.42..4127248.31 rows=12702676 width=200)
> Hash Cond: ("outer".cus_num = "inner".cus_nbr)
> -> Seq Scan on bcp_ddw_ck_cus b (cost=0.00..195690.76 rows=12702676
> width=16)
> -> Hash (cost=874854.34..874854.34 rows=12880834 width=192)
> -> Seq Scan on cdm_ddw_customer (cost=0.00..874854.34
> rows=12880834 width=192)
Yipes, that's a bit of a large hash table, if the planner's estimates
are on-target. What do you have work_mem (sort_mem if pre 8.0) set to,
and how does that compare to actual available RAM? I'm thinking you
might have set work_mem too large and the thing is now swap-thrashing.
regards, tom lane