al_nunes@atua.com.br wrote:
Citando Rod Taylor <pg@rbt.ca>:
Please send an explain analyze from both.
I'm sendin three explains. In the first the Dell machine didn't use existing
indexes, so I turn enable_seqscan off (this is the second explain). The total
cost decreased, but the total time not. The third explain refers to the cheaper
(and faster) machine. The last thing is the query itself.
Nested Loop (cost=9008.68..13596.97 rows=1 width=317) (actual
time=9272.803..65287.304 rows=2604 loops=1)Nested Loop (cost=5155.51..19320.20 rows=1 width=317) (actual
time=480.311..62530.121 rows=2604 loops=1)Hash Join (cost=2.23..11191.77 rows=9 width=134) (actual
time=341.708..21868.167 rows=2604 loops=1)
Well the plan is completely different on the dev machine. Therefore either the PG version or the postgresql.conf is different. No other possible answer.
M