From: =?utf-8?Q?Micha=C5=82_K=C5=82eczek?= <michal@kleczek.org>
Content-Type: text/plain;
	charset=us-ascii
Content-Transfer-Encoding: quoted-printable
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3776.700.51\))
Subject: Partitionwise aggregate and runtime partition pruning
Message-Id: <F54D9FEE-0A39-490A-801F-B2858B527D01@kleczek.org>
Date: Thu, 26 Sep 2024 09:58:18 +0200
To: PG-General Mailing List <pgsql-general@postgresql.org>
Archived-At: <https://www.postgresql.org/message-id/F54D9FEE-0A39-490A-801F-B2858B527D01%40kleczek.org>
Precedence: bulk

Hi All,

I have a question about partition pruning.

Does runtime partition pruning (ie. pruning performed during execution) =
work with partition wise aggregates?

1) I have a setup with a mix of foreign (postgres_fdw) and local =
partitions.
2) I want to perform an aggregate query and I want the aggregates to be =
pushed down to remote servers.

To make it possible I set enable_partitionwise_aggregate to on.

My observation is that partition pruning works during planning time
But with plan_cache_mode =3D force_generic_plan, explain analyse execute =
prepared_stmt(params) shows that
_all_ partitions are scanned (and queries sent to remote severs for all =
remote partitions).

When I set enable_hashagg to false runtime partition pruning works but =
no pushdown is taking place.

Am I missing something or runtime partition pruning is not going to work =
for generic plans and partitionwise aggregates?

Thanks,
Michal