On 2019-Jun-07, Robert Haas wrote:
> Yeah, I wondered whether SnapshotNonVacuumable might've been added
> later, but I was too lazy to check the commit log. I'll try coding up
> that approach and see how it looks.
Thanks.
> But do you have any comment on the question of whether this function
> is actually safe with < ShareLock, per the comments about caching
> HOT-related state across buffer lock releases?
Well, as far as I understand we do hold a buffer pin on the page the
whole time until we abandon it, which prevents HOT pruning, so the root
offset cache should be safe (since heap_page_prune requires cleanup
lock). The thing we don't keep held is a buffer lock, so I/U/D could
occur, but those are not supposed to be hazards for the BRIN use, since
that's covered by the anyvisible / SnapshotNonVacuumable
hack^Wtechnique.
--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services