> sort_mem = 2048
2 meg sort_mem seems on the small side.
> Yes I have an index on actvars.product_level and an index on
> prodlevel.code_level.Both indices have character(12) data types.
Can you force an indexscan to be chosen by setting enable_seqscan off?
If so, what does the explain look like?
BTW, it's always much more useful to show EXPLAIN ANALYZE output than
plain EXPLAIN. The issue is generally "why did the planner misestimate"
and so knowing how its estimates diverge from reality is always a
critical bit of information.
regards, tom lane
In response to
pgsql-performance by date
|Next:||From: Mary Edie Meredith||Date: 2003-09-26 16:11:37|
|Subject: Re: TPC-R benchmarks|
|Previous:||From: Ron Johnson||Date: 2003-09-26 12:37:35|
|Subject: Re: populate table with large csv file|