Re: Adding SMGR discriminator to buffer tags

On Wed, Jul 17, 2019 at 03:01:47PM +1200, Thomas Munro wrote:
> On Tue, Jul 16, 2019 at 10:49 AM Thomas Munro <thomas(dot)munro(at)gmail(dot)com> wrote:
> > I'll go and commit the simple refactoring bits of this work, which
> > just move some stuff belonging to md.c out of smgr.c (see attached).
>
> Pushed. The rest of that earlier patch set is hereby abandoned (at
> least for now). I'll be posting a new-and-improved undo log patch set
> soon, now a couple of patches smaller but back to magic database 9. I
> think I'll probably do that with a new catalog header file that
> defines pseudo-database OIDs.

One suggestion, let's expose the magic oids via a dedicated catalog
pg_smgr so that they can be reserved and accounted for via the scripts
as discussed in [1]. There were suggestions in the thread to use pg_am,
but with the revised pg_am [2], it seems we will be stretching the
meaning of access methods quite a bit, in my opinion, incorrectly.
The benefit of having a dedicated catalog is that we can expose data
particular to smgrs that do not fit in the access methods scope.

[1]
https://www.postgresql.org/message-id/20180821184835.GA1032%4060f81dc409fc.ant.amazon.com
[2] https://www.postgresql.org/docs/devel/catalog-pg-am.html

--
Shawn Debnath
Amazon Web Services (AWS)