On Mon, Jul 31, 2017 at 1:54 PM, Peter Geoghegan <pg@bowt.ie> wrote:
> That is hard to justify. I don't think that failing to set LP_DEAD hints
> is the cost that must be paid to realize a benefit elsewhere, though. I
> don't see much problem with having both benefits consistently. It's
> actually very unlikely that VACUUM will run, and a TID will be recycled
> at exactly the wrong time. We could probably come up with a more
> discriminating way of detecting that that may have happened, at least
> for Postgres 11. We'd continue to use LSN; the slow path would be taken
> when the LSN changed, but we do not give up on setting LP_DEAD bits. I
> think we can justify going to the heap again in this slow path, if
> that's what it takes.
Yeah, that might possibly be a good approach.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company