On Sat, Oct 6, 2018 at 12:17 AM John Naylor <jcnaylor@gmail.com> wrote:
>
> -There'll need to be some performance testing to make sure there's no
> regression, and to choose a good value for the threshold. I'll look
> into that, but if anyone has any ideas for tests, that'll help this
> effort along.
>
Can you try with a Copy command which copies just enough tuples to
fill the pages equivalent to HEAP_FSM_EXTENSION_THRESHOLD? It seems
to me in such a case patch will try each of the blocks multiple times.
It looks quite lame that we have to try again and again the blocks
which we have just filled by ourselves but may be that doesn't matter
much as the threshold value is small.
--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com