| From: | "Jelte Fennema-Nio" <postgres(at)jeltef(dot)nl> |
|---|---|
| To: | "Rui Zhao" <zhaorui126(at)gmail(dot)com> |
| Cc: | "PostgreSQL-development" <pgsql-hackers(at)postgresql(dot)org>, "Etsuro Fujita" <etsuro(dot)fujita(at)gmail(dot)com>, "Robert Haas" <robertmhaas(at)gmail(dot)com>, "Ashutosh Bapat" <ashutosh(dot)bapat(dot)oss(at)gmail(dot)com>, "Narayanan Venkateswaran" <narayananvpostgres(at)gmail(dot)com> |
| Subject: | Re: postgres_fdw: Fix costing of remote sorts without remote estimates |
| Date: | 2026-11-04 22:35:10 |
| Message-ID: | DMMS9RAQOT0Y.28E2MC9GWPYTY@jeltef.nl |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Thu, 8 Oct 2026 at 18:13, Rui Zhao <zhaorui126(at)gmail(dot)com> wrote:
> I found two remaining aggregate-costing cases in 0003.
Both are real inaccuracies, but the planner costs a local Agg the same
way (see create_agg_path()), so fixing them only in postgres_fdw would
make the plans worse. Attached v3 adds 0005 with comments explaining
that.
Also they are pre-existing inaccuracies, they are not introduced by this
patchset.
| Attachment | Content-Type | Size |
|---|---|---|
| v3-0001-postgres_fdw-Track-run-cost-instead-of-total-cost.patch | text/x-patch | 9.2 KB |
| v3-0002-postgres_fdw-Add-tests-for-costing-local-work-bel.patch | text/x-patch | 6.4 KB |
| v3-0003-postgres_fdw-Separate-remote-transfer-and-local-c.patch | text/x-patch | 17.5 KB |
| v3-0004-postgres_fdw-Fix-costing-of-remote-sorts-without-.patch | text/x-patch | 34.8 KB |
| v3-0005-postgres_fdw-Explain-where-remote-tlist-costing-d.patch | text/x-patch | 2.9 KB |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Heikki Linnakangas | 2026-11-04 22:45:25 | Re: enhancing pg_basebackup speeds up to ~23Gbps (small fixes + io_uring/Direct I/O) |
| Previous Message | Michael Paquier | 2026-11-04 22:22:03 | Re: Fix cumulative WAL statistics double-counting from parallel workers |