Tom Lane wrote:
Steve Clark <sclark@xxxxxxxxxxxxx> writes:
explain shows:
Aggregate (cost=4712921585.30..4712921585.31 rows=1 width=0)
-> Seq Scan on t_event_ack_log a (cost=103170.29..4712920878.60
rows=282677 width=0)
Filter: (NOT (subplan))
SubPlan
-> Materialize (cost=103170.29..117301.92 rows=1016163
width=4)
-> Index Scan using pk_tuel_eln on t_unit_event_log
(cost=0.00..98184.12 rows=1016163 width=4)
Yeah, that's going to suck. A brute force solution is to see if you
can get it to switch to a "hashed subplan" by increasing work_mem.
Also, whatever is the ORDER BY for?
regards, tom lane
without the order by it wants to do a seq scan of t_unit_event_log.
see below:
explain select count(*) from t_event_ack_log where event_log_no not
in (select event_log_no from t_unit_event_log);
QUERY PLAN
----------------------------------------------------------------------------------------
Aggregate (cost=12144872193.82..12144872193.82 rows=1 width=0)
-> Seq Scan on t_event_ack_log (cost=0.00..12144871485.07
rows=283497 width=0)
Filter: (NOT (subplan))
SubPlan
-> Seq Scan on t_unit_event_log (cost=0.00..40286.56
rows=1021156 width=4)
(5 rows)
Will try increasing work_memory.
---------------------------(end of broadcast)---------------------------
TIP 1: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo@xxxxxxxxxxxxxx so that your
message can get through to the mailing list cleanly
