Re: pg_upgrade and toasted pg_largeobject

Robert Haas wrote:
> On Tue, May 3, 2016 at 2:09 PM, Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> wrote:
> > Tom Lane wrote:
> >> Any thoughts what to do with this? We could decide that it's a bug fix
> >> and backpatch, or decide that it's a new feature and delay till 9.7,
> >> or decide that it's a minor bug fix and add it to 9.6 only. I kinda lean
> >> towards the last alternative.
> >
> > How about backpatching patch 1 all the way back, and putting the others
> > in 9.6?
>
> Why would we do that? It seems very odd to back-patch a pure
> refactoring - isn't that taking a risk for no benefit?

From Tom's description, what is there works by chance only, and maybe
not even in all cases. The rest of the patches are to fix one
particular problem, which perhaps is not overly serious, but maybe some
other future problem will be discovered and we will want to have patch 1
installed.

My inclination is actually to put the whole series back to 9.2, but if
we don't want to do that, then backpatching just the first one seems to
make pg_upgrade more amenable to future bugfixes.

(I say "back to 9.2" instead of "back to 9.1" because surely we don't
care all that much about upgrades *to* 9.1, since it's going unsupported
soon.)

--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services