Date: Fri, 7 Jan 2022 10:09:21 -0600
From: Justin Pryzby <pryzby@telsasoft.com>
To: "Finnerty, Jim" <jfinnert@amazon.com>
Cc: Alexander Korotkov <aekorotkov@gmail.com>,
	Maxim Orlov <orlovmg@gmail.com>, pgsql-hackers@postgresql.org
Subject: Re: Add 64-bit XIDs into PostgreSQL 15
Message-ID: <20220107160921.GD14051@telsasoft.com>
References: 
 <CACG=ezZe1NQSCnfHOr78AtAZxJZeCvxrts0ygrxYwe=pyyjVWA@mail.gmail.com>
 <20220104193220.GL15820@tamriel.snowman.net>
 <CB919F14-2B98-4CD3-84C0-91945E9E7EF5@amazon.com>
 <6960BD21-E46F-4682-9A59-17716B346019@amazon.com>
 <CAPpHfdvyu=__Le5wYxCiOQtVWoHgteEAw+zXOLseiBpFdhPdFg@mail.gmail.com>
 <9AFAF000-2E9E-443E-9EC3-7B67D6A03E5B@amazon.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <9AFAF000-2E9E-443E-9EC3-7B67D6A03E5B@amazon.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Archived-At: 
 <https://www.postgresql.org/message-id/20220107160921.GD14051%40telsasoft.com>
Precedence: bulk

On Fri, Jan 07, 2022 at 03:53:51PM +0000, Finnerty, Jim wrote:
> I'd still like a plan to retire the "double xmax" representation eventually.  Previously I suggested that this could be done as a post-process, before upgrade is complete, but that could potentially make upgrade very slow. 
> 
> Another way to retire the "double xmax" representation eventually could be to disallow "double xmax" pages in subsequent major version upgrades (e.g. to PG16, if "double xmax" pages are introduced in PG15).  This gives the luxury of time after a fast upgrade to convert all pages to contain the epochs, while still providing a path to more maintainable code in the future.

Yes, but how are you planning to rewrite it?  Is vacuum enough?
I suppose it'd need FREEZE + DISABLE_PAGE_SKIPPING ?

This would preclude upgrading "across" v15.  Maybe that'd be okay, but it'd be
a new and atypical restriction.

How would you enforce that it'd been run on v15 before upgrading to pg16 ?

You'd need to track whether vacuum had completed the necessary steps in pg15.
I don't think it'd be okay to make pg_upgrade --check to read every tuple.

The "keeping track" part is what reminds me of the online checksum patch.
It'd be ideal if there were a generic solution to this kind of task, or at
least a "model" process to follow.

-- 
Justin