Re: [DESIGN] ParallelAppend

On 17 November 2015 at 20:08, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> On Tue, Nov 17, 2015 at 4:26 AM, Thom Brown <thom(at)linux(dot)com> wrote:
>
>> However, the first parallel seq scan shows it getting 170314 rows.
>> Another run shows it getting 194165 rows. The final result is
>> correct, but as you can see from the rows on the Append node (59094295
>> rows), it doesn't match the number of rows on the Gather node
>> (30000000).
>
> Is this the same issue reported in
> http://www.postgresql.org/message-id/CAFj8pRBF-i=qDg9b5nZrXYfChzBEZWmthxYPhidQvwoMOjHtzg@mail.gmail.com
> and not yet fixed? I am inclined to think it probably is.

Yes, that seems to be the same issue.

>> And also, for some reason, I can no longer get this using more than 2
>> workers, even with max_worker_processes = 16 and max_parallel_degree =
>> 12. I don't know if that's anything to do with this patch though.
>
> The number of workers is limited based on the size of the largest
> table involved in the Append. That probably needs considerable
> improvement, of course, but this patch is still a step forward over
> not-this-patch.

Ah, okay. I wasn't aware of this. I'll bear that in mind in future.

Thom