"Andrew Bartley" <abartley@evolvosystems.com> writes:
> I have tried to but together a shell script to touch these files. But it
> seems that the table in question has a second pg_toast oid associated with
> it.
That's interesting ... if both those physical files went away while the
pg_class rows remained, it suggests some systematic problem rather than
just a random glitch. I doubt we'll be able to learn anything about the
cause if we're not allowed into the database machine though :-(
Anyway, the pg_class.reltoastrelid field for the temp table gives you
the OID of the pg_class row for its TOAST table, from which you can get
the relfilenode value to touch for the TOAST table.
Very likely, the TOAST table's index has got the same disease: to find
it, get the reltoastidxid field from the TOAST table's pg_class row, and
again go to that OID in pg_class to see its relfilenode value.
(Actually, in 7.2 you'll almost certainly find that relfilenode of each
of these rows matches its OID, but to be perfectly safe you should look
up and use the relfilenode.)
regards, tom lane