Re: explain HashAggregate to report bucket and memory stats

Hi,

On 2020-03-13 10:53:17 -0700, Jeff Davis wrote:
> On Fri, 2020-03-13 at 10:27 -0700, Andres Freund wrote:
> > On 2020-03-13 10:15:46 -0700, Jeff Davis wrote:
> > > Also, is there a reason you report two different memory values
> > > (hashtable and tuples)? I don't object, but it seems like a little
> > > too
> > > much detail.
> >
> > Seems useful to me - the hashtable is pre-allocated based on
> > estimates,
> > whereas the tuples are allocated "on demand". So seeing the
> > difference
> > will allow to investigate the more crucial issue...
>
> Then do we also want to report separately on the by-ref transition
> values? That could be useful if you are using ARRAY_AGG and the states
> grow larger than you might expect.

I can see that being valuable - I've had to debug cases with too much
memory being used due to aggregate transitions before. Right now it'd be
mixed in with tuples, I believe - and we'd need a separate context for
tracking the transition values? Due to that I'm inclined to not report
separately for now.

Greetings,

Andres Freund