Re: pg_dump, MVCC and consistency

On Mon, Oct 24, 2005 at 11:25:00AM -0600, Michael Fuhr wrote:
> PostgreSQL 8.1 makes checks to avoid data loss due to transaction
> ID wraparound, but there's one situation I'm not sure how it handles:
> when a transaction is so long-lived that it would appear to be in
> the future of newly-created transactions due to wraparound. I'd
> have to dig into the source code to find out if that's possible,
> and if so, what happens. Maybe one of the developers will comment.

To avoid this you need to do a VACUUM FULL over the database at least
once every two billion transactions (not statements or tuples,
transactions). To that end, the server begins complaining after one
billion. I've never seen this in practice. Perhaps you could calculate
how long it would take to do that many transactions. Most systems will
never see it...

Hope this helps,
--
Martijn van Oosterhout <kleptog(at)svana(dot)org> http://svana.org/kleptog/
> Patent. n. Genius is 5% inspiration and 95% perspiration. A patent is a
> tool for doing 5% of the work and then sitting around waiting for someone
> else to do the other 95% so you can sue them.