> No, explain analyze for the query that wouldn't execute before but now
> does, with, I assume, a large work_mem. I'd like to see how it
> differes from the one with smaller work_mem.
Ah, I pasted that in an earlier email, sent February 10, 2009 9:58:00 AM
GMT+13:00... that plan was the one using still the 128MB of work_mem after
changing the overcommit_ratio to 80.
Regards,
Matt