In accord with Murphy's Fifth Postulate, after being driven to send a
message to pgsql-admin, I have since solved the problem. Thus, I send this
message to record the answers for posterity.
At 12:06 AM 4/5/00 -0700, Charles Martin wrote:
>There are at least three major problems:
>
> * VACUUM ANALYZE and pg_dump fail on one of our tables.
There were two bogus rows in the "users" table. One with OID 0 and one
with OID 1634628924; how these rows got into the table I have no idea, but
they were filled with bogus data. Deleting these rows and vacuuming fixed
the problems with this table (though I still see "no parent tuple" messages).
* Thousands of spurious files have been created in the
> data/base/<dbname> directory.
I just deleted all these spurious files. (They were all of zero
length.) I have no idea why they were created in the first place.
Now that I can pg_dump, time for PostgreSQL 7.0 beta 2!