On Sun, Dec 9, 2018 at 6:24 AM Justin Pryzby <pryzby@telsasoft.com> wrote:
>
> The docs say:
> https://www.postgresql.org/docs/current/runtime-config-query.html
> |min_parallel_table_scan_size Sets the minimum amount of table data that must be scanned in order for a parallel scan
tobe considered. [...]
>
> I'd like to set parallel_min_table_size=32MB, but it looks like that won't do
> what I intend for at least one of our tables using inheritence.
>
> It seems to me that an individual table should not be scanned in parallel if
> its size is below the threshold, even if it's a child and has siblings which
> are larger and scanned in parallel.
>
> I found that the current behavior seems to be more or less deliberate,
Yes, you are right.
> but
> maybe should be revisited following implementation of "parallel append" node,
> as previously discussed.
>
Why? How does parallel append help for individual table/partition
scan? If you want the parallel scan to be forced for any individual
table, you might want to try setting 'parallel_workers' parameter for
that table. For example 'Alter Table tbl Set (parallel_workers=2);'
--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com