On Sat, Jun 30, 2018 at 4:08 AM, Alvaro Herrera
<alvherre@2ndquadrant.com> wrote:
> On 2018-Jun-29, Tom Lane wrote:
>
>> Alvaro Herrera <alvherre@2ndquadrant.com> writes:
>> > I'm CCing Tom here, as author of the patch that caused (most of) the
>> > issue.
>>
>> Uh ... me? I thought this was a parallel-query issue, which I've
>> pretty much not been involved in.
>
> Well, maybe it's a matter of opinion. Amit K said a few messages back
> that your 01edb5c7fc3b ("Improve division of labor between
> execParallel.c and nodeGather[Merge].c.") had changed the way these
> numbers are printed, but only now I realize that he then indicated that
> a different code path was already behaving in that way.
>
> I stand by my opinion that we should not give misleading/confusing info;
> either let's show it all in the default output, or only do it in
> VERBOSE, but if the latter then let's suppress the misleading numbers in
> the default output.
>
It is not clear to me what exactly is your expectation. Can you be a
bit more specific? AFAIU, the primary confusion to OP is that the
aggregated stats like buffer_usage (Buffers: ..) should be displayed
correctly at Gather/Gather Merge and nodes above it. Currently, those
are being displayed correctly at Parallel (Seq Scan) nodes.
--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com