On Oct 7, 2006, at 6:41 PM, Chris Browne wrote:
> This could also be a situation where adding a few useful indexes might
> fix a lot of ills. Better to try to help fix the problems so as to
> help show that the comparisons are way off base rather than to simply
> cast stones...
i'm too tight for cash to afford being wrong right now...
but I'd otherwise bet that the issue was from not vacuum analyzing
i've routinely had 3,9,12, i think even a 14 table join that would
take forever to run...
until i realized that i added/dropped an index and forgot to run
analyze. then they all work within a matter of split seconds. all of
them.
i've seen not just dramatic, but drastic , changes in performance and
the planner's output before and after a vacuum analyze of the db.
i'm really confident thats the problem. unfortunately, they have a
max_db contact email, and not a postgres. so i don't know who to
check with to see if they ran it or not.