Re: 64-bit XIDs in deleted nbtree pages

On Sat, Feb 13, 2021 at 10:47 PM Peter Geoghegan <pg(at)bowt(dot)ie> wrote:
> It will be rare. But more importantly, the fact that scenario is now
> an extreme case justifies treating it as an extreme case. We can teach
> _bt_vacuum_needs_cleanup() to recognize it as an extreme case, too. In
> particular, I think that it will now be okay to increase the threshold
> applied when considering deleted pages inside
> _bt_vacuum_needs_cleanup(). It was 2.5% of the index size in v3 of the
> patch. But in v4, which has the new recycling enhancement, I think
> that it would be sensible to make it 5%, or maybe even 10%. This
> naturally makes Masahiko's problem scenario unlikely to actually
> result in a truly wasted call to btvacuumscan().

Attached is v4, which has the "recycle pages that we ourselves deleted
during this same VACUUM operation" enhancement. It also doubles the
_bt_vacuum_needs_cleanup() threshold applied to deleted pages -- it
goes from 2.5% to 5%. The new patch is the patch series (v4-0002-*)
certainly needs more polishing. I'm posting what I have now because v3
has bitrot.

Benchmarking has shown that the enhancement in v4-0002-* can
significantly reduce the amount of index bloat in two of the
BenchmarkSQL/TPC-C indexes.

--
Peter Geoghegan