On Mon, 3 May 2021 at 16:39, Matthias van de Meent
<boekewurm+postgres@gmail.com> wrote:
> I am planning on fixing this patch sometime
> before the next commit fest so that we can truncate the LP array
> during hot pruning as well, instead of only doing so in the 2nd VACUUM
> pass.
PFA the updated version of this patch. Apart from adding line pointer
truncation in PageRepairFragmentation (as in the earlier patches), I
also altered PageTruncateLinePointerArray to clean up all trailing
line pointers, even if it was the last item on the page.
This means that for 32-bit systems, pages that have once had tuples
(but have been cleared since) can now be used again for
MaxHeapTupleSize insertions. Without this patch, an emptied page would
always have at least one line pointer left, which equates to
MaxHeapTupleSize actual free space, but PageGetFreeSpace always
subtracts sizeof(ItemIdData), leaving the perceived free space as
reported to the FSM less than MaxHeapTupleSize if the page has any
line pointers.
For 64-bit systems, this is not as much of a problem, because
MaxHeapTupleSize is 4 bytes smaller on those systems, which leaves us
with 1 line pointer as margin for the FSM to recognise the page as
free enough for one MaxHeapTupleSize item.
With regards,
Matthias van de Meent
