| From: | Michael Paquier <michael(at)paquier(dot)xyz> |
|---|---|
| To: | Nitin Jadhav <nitinjadhavpostgres(at)gmail(dot)com> |
| Cc: | Adam Lee <adam8157(at)gmail(dot)com>, Heikki Linnakangas <hlinnaka(at)iki(dot)fi>, pgsql-hackers(at)lists(dot)postgresql(dot)org |
| Subject: | Re: [PATCH] Fix minRecoveryPoint not advanced past checkpoint in CreateRestartPoint |
| Date: | 2026-06-10 01:16:25 |
| Message-ID: | aii66Z4j3W2gm8K1@paquier.xyz |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Tue, Jun 09, 2026 at 06:08:36PM +0530, Nitin Jadhav wrote:
> Given that the current behavior creates inconsistency across
> recovery_target_action settings (pause and promote behave differently
> than shutdown), external tools such as pg_rewind, backup solutions,
> and monitoring systems depend on this value for operational decisions,
> I support moving forward with this patch.
FWIW, I am still unconvinced even after a second read of the thread.
In basically all workloads (hand-waving a number but let's say 99%),
we are going to have page flushes anyway between a CHECKPOINT record
replayed and the recovery target, where each one is going to naturally
update the minRecoveryPoint. I'd still want less paths that update
minRecoveryPoint at the end of the day, not more of them.
--
Michael
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Michael Paquier | 2026-06-10 01:18:40 | Re: Fix DROP PROPERTY GRAPH "unsupported object class" error |
| Previous Message | Enrique Sánchez | 2026-06-10 00:57:29 | Re: Extended statistics improvement: multi-column MCV missing values |