On Mar 8, 2010, at 7:16 AM, Tom Lane wrote:
> After it's done you probably ought to look into your autovacuum
> settings. You shouldn't have gotten into a state with so many dead
> tuples in the first place, I would think, if autovac were functioning
> properly.
Yeah, it turns out our autovacuums kept getting canceled because this table gets a lot of churn. Now that we've turn off the cost delay, things are much happier. Thanks!