Re: data volume resizing and large objects

Tom

You hit the nail on the head actually! Further delving into the issue
revealed that but I hadnt had a chance to post.

We recently used copy to put some lobs back into the database, about 70k of
them. All of these lobs were assigned sequential oids incremented by 1. We
then wrapped around and ended up in that block of oids again.

The irony of all of this is we have been trying to migrate to 8.1 for months
now but have not been able to do so due to infrastructure issues beyond our
control.

Is there anyway to force 8.0 to start at a specific value? That would at
least get us back up and running.

Thanks

-bill

On Jan 4, 2008 3:42 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> "beer(at)cmu(dot)edu" <beer(at)cmu(dot)edu> writes:
> > lo_creat fails with:
> > ERROR: large object <oid> already exists
>
> Well, does it?
>
> What PG version is this? I'm suspicious that it's pre-8.1 and the
> problem is that your OID counter has wrapped around to a region that
> is already pretty densely populated with large objects. 8.1 and up
> are able to pick unused OIDs but earlier versions just fail on
> collisions.
>
> regards, tom lane
>