On Wed, Jun 16, 2021 at 9:03 AM Peter Geoghegan <pg@bowt.ie> wrote:
> On Wed, Jun 16, 2021 at 3:59 AM Matthias van de Meent
> > So the implicit assumption in heap_page_prune that
> > HeapTupleSatisfiesVacuum(OldestXmin) is always consistent with
> > heap_prune_satisfies_vacuum(vacrel) has never been true. In that case,
> > we'll need to redo the condition in heap_page_prune as well.
>
> I don't think that this shows that the assumption within
> lazy_scan_prune() (the assumption that both "satisfies vacuum"
> functions agree) is wrong, with the obvious exception of cases
> involving the bug that Justin reported. GlobalVis*.maybe_needed is
> supposed to be conservative.
I suppose it's true that they can disagree because we call
vacuum_set_xid_limits() to get an OldestXmin inside vacuumlazy.c
before calling GlobalVisTestFor() inside vacuumlazy.c to get a
vistest. But that only implies that a tuple that would have been
considered RECENTLY_DEAD inside lazy_scan_prune() (it just missed
being considered DEAD according to OldestXmin) is seen as an LP_DEAD
stub line pointer. Which really means it's DEAD to lazy_scan_prune()
anyway. These days the only way that lazy_scan_prune() can consider a
tuple fully DEAD is if it's no longer a tuple -- it has to actually be
an LP_DEAD stub line pointer.
It's really no different to an opportunistic prune that concurrently
prunes tuples that VACUUM would have seen as RECENTLY_DEAD if it was
going solely on the OldestXmin cutoff. There are certain kinds of
tables where non-HOT updates and opportunistic pruning constantly
leave behind loads of LP_DEAD items. Pruning inside VACUUM won't do
much of the total required pruning at all. That'll mean that some
DEAD/LP_DEAD items will become dead long after a VACUUM starts, while
nevertheless being removed by the same VACUUM. Of course there is no
way for lazy_scan_prune() to distinguish one LP_DEAD item from another
-- they're all stubs without tuple storage, and without a tuple header
with XIDs.
--
Peter Geoghegan