Horton

Re: Default setting for enable_hashagg_disk

Peter Geoghegan <pg@bowt.ie>

View thread View on postgr.es
From: Peter Geoghegan <pg@bowt.ie>
To: Jeff Davis <pgsql@j-davis.com>
Cc: Alvaro Herrera <alvherre@2ndquadrant.com>, David Rowley <dgrowleyml@gmail.com>, Stephen Frost <sfrost@snowman.net>, Tom Lane <tgl@sss.pgh.pa.us>, Andres Freund <andres@anarazel.de>, Bruce Momjian <bruce@momjian.us>, Robert Haas <robertmhaas@gmail.com>, Justin Pryzby <pryzby@telsasoft.com>, Melanie Plageman <melanieplageman@gmail.com>, Tomas Vondra <tomas.vondra@2ndquadrant.com>, "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Date: 2020-07-23T23:32:07Z
Lists: pgsql-hackers, pgsql-docs

Attachments

On Fri, Jul 17, 2020 at 5:13 PM Jeff Davis <pgsql@j-davis.com> wrote:
> The patch itself looks reasonable to me. I don't see a lot of obvious
> dangers, but perhaps someone would like to take a closer look at the
> planner changes as you suggest.

Attached is v3 of the hash_mem_multiplier patch series, which now has
a preparatory patch that removes hashagg_avoid_disk_plan. What do you
think of this approach, Jeff?

It seems as if removing hashagg_avoid_disk_plan will necessitate
removing various old bits of planner.c that were concerned with
avoiding hash aggs that spill (the bits that hashagg_avoid_disk_plan
skips in the common case where it's turned off). This makes v3-0001-*
a bit trickier than I imagined it would have to be. At least it lowers
the footprint of the hash_mem_multiplier code added by v3-0002-*
(compared to the last version of the patch).

I find the partial group paths stuff added to planner.c by commit
4f15e5d09de rather confusing (that commit was preparatory work for the
main feature commit e2f1eb0e). Hopefully the
hash_mem_multiplier-removal patch didn't get anything wrong in this
area. Perhaps Robert can comment on this as the committer of record
for partition-wise grouping/aggregation.

I would like to commit this patch series by next week, and close out
the two relevant open items. Separately, I suspect that we'll also
need to update the cost model for hash aggs that spill, but that now
seems like a totally unrelated matter. I'm waiting to hear back from
Tomas about that. Tomas?

Thanks
--
Peter Geoghegan

Commits

  1. Add hash_mem_multiplier GUC.

  2. HashAgg: use better cardinality estimate for recursive spilling.

  3. Remove hashagg_avoid_disk_plan GUC.

  4. Doc fixup for hashagg_avoid_disk_plan GUC.

  5. Rework HashAgg GUCs.

  6. Disk-based Hash Aggregation.

  7. Implement partition-wise grouping/aggregation.

  8. Defer creation of partially-grouped relation until it's needed.