> Is this xmin quite a bit older than what you get for a freshly-created
> temp table?
I would say yes, this is a test system which is highly stressed from
time to time, but does not get continuous load.
test03=> select oid, xmin from pg_type where typname = 'temp_report';
oid | xmin
----------+-----------
58293995 | 220215039
(1 row)
test03=> create temp table test_new_temp_table (a text);
CREATE TABLE
test03=> select xmin from pg_type where typname = 'test_new_temp_table';
xmin
-----------
236080536
(1 row)
> Can you find any entries in pg_depend that show the above
> OID as either objid or refobjid?
No:
test03=> select * from pg_depend where objid = 58293995 or refobjid =
58293995;
classid | objid | objsubid | refclassid | refobjid | refobjsubid |
deptype
---------+-------+----------+------------+----------+-------------+---------
(0 rows)
> Also, it's fairly likely (not certain) that the table associated with
> this pg_type entry had OID one less, ie 58293994. Can you find any
> trace of that OID in pg_depend,
No trace. I executed:
test03=> select count(*) from pg_depend where classid between 58293995 -
100 and 58293995 + 100;
count
-------
0
(1 row)
... and then with each of (objid objsubid refclassid refobjid
refobjsubid deptype) instead of classid, with the same result.
> ... or in pg_class for that matter?
test03=> select count(*) from pg_class where oid between 58293995 - 100
and 58293995 + 100;
count
-------
0
(1 row)
Any other place to check ?
Cheers,
Csaba.