nolan@celery.tssi.com writes:
> Is there a documented problem with optimizing integer key fields
> in 7.3.3?
I see nothing particularly wrong with that query plan.
> When I try to do an update, it takes several hours.
That seems excessive for a 700K-row update, but I doubt that the query
plan is the problem. I'm betting there's a lot of per-update overhead
due to triggers or foreign keys. What have you got in that line? Do
you have indexes on both sides of any foreign-key relationships that
missing_ids participates in?
regards, tom lane