Re: Pg_upgrade and toast tables bug discovered

From: Bruce Momjian <bruce(at)momjian(dot)us>
To: Andres Freund <andres(at)2ndquadrant(dot)com>
Cc: Robert Haas <robertmhaas(at)gmail(dot)com>, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Pg_upgrade and toast tables bug discovered
Date: 2014-07-10 22:17:14
Message-ID: 20140710221714.GC7366@momjian.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Thu, Jul 10, 2014 at 11:11:19PM +0200, Andres Freund wrote:
> On 2014-07-10 16:33:40 -0400, Bruce Momjian wrote:
> > On Thu, Jul 10, 2014 at 10:46:30AM -0400, Robert Haas wrote:
> > > > Agreed. I am now thinking we could harness the code that already exists
> > > > to optionally add a TOAST table as part of ALTER TABLE ADD COLUMN. We
> > > > would just need an entry point to call it from pg_upgrade, either via an
> > > > SQL command that checks (and hopefully doesn't do anything else), or a C
> > > > function that does it, e.g. VACUUM would be trivial to run on every
> > > > database, but I don't think it tests that; is _could_ in binary_upgrade
> > > > mode. However, the idea of having a C function plug into the guts of
> > > > the server and call internal functions makes me uncomforable.
> > >
> > > Well, pg_upgrade_support's charter is basically to provide access to
> > > the guts of the server in ways we wouldn't normally allow; all that
> > > next-OID stuff is basically exactly that. So I don't think this is
> > > such a big deal. It needs to be properly commented, of course.
> >
> > If you look at how oid assignment is handled, it is done in a very
> > surgical way, i.e. pg_upgrade_support sets a global variable, and the
> > variable triggers different behavior in a CREATE command. This change
> > would be far more invasive than that.
>
> Meh. It's only somewhat surgical because there's pg_upgrade specific
> code sprinkled in the backend at strategic places. That's the contrary
> of a clear abstraction barrier. And arguably a function call from a SQL
> C function has a much clearer abstraction barrier.

Well, we are going to need to call internal C functions, often bypassing
their typical call sites and the assumption about locking, etc. Perhaps
this could be done from a plpgsql function. We could add and drop a
dummy column to force TOAST table creation --- we would then only need a
way to detect if a function _needs_ a TOAST table, which was skipped in
binary upgrade mode previously.

That might be a minimalistic approach.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ Everyone has their own god. +

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Jeff Janes 2014-07-10 22:17:50 Re: Minmax indexes
Previous Message Vik Fearing 2014-07-10 22:01:10 Re: Missing autocomplete for CREATE DATABASE