On Mon, 2005-10-17 at 12:04, Chris Browne wrote:
> jeff@endpoint.com (Jeff Boes) writes:
> > On behalf of a previous employer who, due to major downsizing, is left
> > without an answer to this:
> >
> > What's the best way to treat impending OID wrap-around in a 7.4.8
> > database? This DB has been online for about 18 months, and is expected
> > to hit the dreaded wrap-around in about a month. At an application
> > level, there's some small chance that OID wrap will cause failures:
> > there's code that writes rows, and (for tables without primary keys)
> > relies on DBD::Pg's pg_oid_status method to identify the last INSERT.
> > (The vast majority of the code does not rely on this, but there's a tiny
> > fraction that does, and rewriting the code to remove this is not an
> > option due to the aforementioned downsizing.)
> >
> > My immediate take on this is that the only sure way to avoid the
> > problem would be to dump and reload the database. Due to the size of
> > the database, this is likely to take most of a day. Any other
> > options?
>
> Another method would be to use replication to load the data into a
> fresh DB instance.
>
> You could, in principle, use Slony-I to do this; that is an intended
> sort of "use case."
I thought that with 7.4 all you needed was to vacuum all databases (full
maybe???) to prevent oid wraparound.