Marie,
> I ran the vacuum for selected tables. It looks fine, I think, but I amn't
> always sure what I am reading in output.
So much for the easy answer. The reason I wanted to see a VACUUM FULL is
that the query on the "bad" database is taking a long time to return even the
first row of many of its sub-parts. This is usually the result of not
running VACUUM FULL after a lot of deletions.
However, your problem apparently is something else. Is is possible that
there is some kind of disk access problem for the bad database copy? Is
there a difference in where its files are physically located?
--
-Josh Berkus
Aglio Database Solutions
San Francisco