On Thu, Jul 19, 2018 at 4:29 AM, Masahiko Sawada <sawada.mshk@gmail.com> wrote:
>> One area that might be worth investigating is retail index tuple
>> deletion performed within the executor in the event of non-HOT
>> updates. Maybe LP_REDIRECT could be repurposed to mean "ghost record",
>> at least in unique index tuples with no NULL values. The idea is that
>> MVCC index scans can skip over those if they've already found a
>> visible tuple with the same value.
>
> I think that's a good idea. The overhead of marking it as ghost seems
> small and it would speed up index scans. If MVCC index scans have
> already found a visible tuples with the same value they can not only
> skip scanning but also kill them? If can, we can kill index tuples
> without checking the heap.
I think you're talking about making LP_REDIRECT marking in nbtree
represent a "recently dead" hint: the deleting transaction has
committed, and so we are 100% sure that the tuple is about to become
garbage, but it cannot be LP_DEAD just yet because it needs to stay
around for the benefit of at least one existing snapshot/long running
transaction.
That's a different idea to what I talked about, since it could perhaps
work in a way that's much closer to LP_DEAD/kill_prior_tuple (no extra
executor stuff is required). I'm not sure if your idea is better or
worse than what I suggested, though. It would certainly be easier to
implement.
--
Peter Geoghegan