| From: | Nathan Bossart <nathandbossart(at)gmail(dot)com> |
|---|---|
| To: | Hannu Krosing <hannuk(at)google(dot)com> |
| Cc: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Nitin Motiani <nitinmotiani(at)google(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
| Subject: | Re: pg_upgrade: transfer pg_largeobject_metadata's files when possible |
| Date: | 2025-08-19 19:18:40 |
| Message-ID: | aKTOEGwYP1QbzE1K@nathan |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Tue, Aug 19, 2025 at 09:49:26AM +0200, Hannu Krosing wrote:
> Have you considered re-creating pg_shdepend from
> pg_largeobject_metadata directly instead of having special cases for
> dumping it ?
I considered it when you last brought up the idea [0], but my testing
indicated that it's measurably slower.
> It would also be useful in cases of old (pg_upgraded since before pg
> 12) databases which might be missing it anyway.
We only use COPY for upgrades from v12 and newer, and the patch at hand
only applies to v16 and newer. There should be no need to repair
pg_shdepend for any such upgrades because we haven't copied/linked the
files since before v12.
[0] https://postgr.es/m/CAMT0RQTXiqH7zdQEVSVd2L7_Cw4wQ1eHOD8hfZ%2B0vecMXJWc-w%40mail.gmail.com
--
nathan
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Masahiko Sawada | 2025-08-19 19:23:44 | Re: Conflict detection for update_deleted in logical replication |
| Previous Message | Masahiko Sawada | 2025-08-19 19:16:47 | Re: Add CHECK_FOR_INTERRUPTS in pg_buffercache_pages while scanning the buffers |