Re: XLog size reductions: Reduced XLog record header size for PG17
Matthias van de Meent <boekewurm+postgres@gmail.com>
From: Matthias van de Meent <boekewurm+postgres@gmail.com>
To: PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Cc: Heikki Linnakangas <hlinnaka@iki.fi>, Robert Haas <robertmhaas@gmail.com>, Dilip Kumar <dilipbalaut@gmail.com>, Andres Freund <andres@anarazel.de>
Date: 2023-07-12T12:50:52Z
Lists: pgsql-hackers
Attachments
- v4-0003-Rename-rmgr_identify-info-bits-argument-to-rmgrin.patch (application/octet-stream) patch v4-0003
- v4-0005-Stop-the-logging-of-XIDs-in-records-where-the-XID.patch (application/octet-stream) patch v4-0005
- v4-0004-Move-rmgr-info-bits-into-their-own-field-in-the-x.patch (application/octet-stream) patch v4-0004
- v4-0002-Replace-accesses-to-rmgr-s-XLogRecord-info-bit-ac.patch (application/octet-stream) patch v4-0002
- v4-0001-Reduce-overhead-of-small-block-data-in-xlog-recor.patch (application/octet-stream) patch v4-0001
- v4-0006-Add-RM_INVALID-an-invalid-resource-manager.patch (application/octet-stream) patch v4-0006
- v4-0007-Reformat-the-XLog-record-header.patch (application/octet-stream) patch v4-0007
On Mon, 3 Jul 2023 at 13:08, Matthias van de Meent <boekewurm+postgres@gmail.com> wrote: > > On Fri, 30 Jun 2023 at 17:36, Matthias van de Meent > <boekewurm+postgres@gmail.com> wrote: > > > > Hi, > > > > The attached v2 patchset contains some small fixes for the failing > > cfbot 32-bit tests - at least locally it does so. > > > > I'd overlooked one remaining use of MAXALIGN64() in xlog.c in the last > > patch of the set, which has now been updated to XLP_ALIGN as well. > > Additionally, XLP_ALIGN has been updated to use TYPEALIGN64 instead of > > TYPEALIGN so that we don't lose bits of the aligned value in 32-bit > > systems. > > Apparently there was some usage of MAXALIGN() in xlogreader that I'd > missed, and which only shows up in TAP tests. In v3 I've fixed that, > together with some improved early detection of invalid record headers. Another fix for CFBot - pg_waldump tests which were added in 96063e28 exposed an issue in my patchset related to RM_INVALID_ID. v4 splits former patch 0006 into two: now 0006 adds RM_INVALID and does the rmgr-related changes in the code, and 0007 does the WAL disk format overhaul. Kind regards, Matthias van de Meent Neon (https://neon.tech/)