Problem with OID wrap

Martin Weinberg <weinberg@osprey.astro.umass.edu> writes:

It seems that OID in one of our databases wrapped during a data
load from tape.
The wrap is quite plausible based on the volume of the database and
the number of tables that have been loaded and reloaded. I would have
thought that an OID wrap would be harmless (except in the loss of
uniqueness and meaning for any query which depended on this).
However, the user complains that part of the table disappeared? Is
this possible?

I'd be wondering about transaction ID wrap, myself. You might care
to consult the discussion of transaction ID wrap in the 7.2 development
docs (under admin guide "routine maintenance", IIRC).

regards, tom lane

Tom Lane wrote on Fri, 14 Dec 2001 10:54:09 EST

Martin Weinberg <weinberg@osprey.astro.umass.edu> writes:

It seems that OID in one of our databases wrapped during a data
load from tape.
The wrap is quite plausible based on the volume of the database and
the number of tables that have been loaded and reloaded. I would have
thought that an OID wrap would be harmless (except in the loss of
uniqueness and meaning for any query which depended on this).
However, the user complains that part of the table disappeared? Is
this possible?

I'd be wondering about transaction ID wrap, myself. You might care
to consult the discussion of transaction ID wrap in the 7.2 development
docs (under admin guide "routine maintenance", IIRC).

Sound like the best thing to do is to move to 7.2beta as soon as
possible. A dump and a re-initdb will be painful on a regular basis.
The 7.2 strategy seems reasonable and should work fine for us.
Ironically, it turns out that we don't use the transactional features
anyway. Is there a way in the mean time to disable this in 7.1.3?

Thanks Tom!

--Martin