Re: pg_upgrade < 9.3 -> >=9.3 misses a step around multixacts

On 2014-07-22 16:08:37 +0700, Stuart Bishop wrote:
> On 21 July 2014 00:37, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>
> > I'm a bit inclined to not say anything about fix procedures in the release
> > notes, because I'm not sure that this is a problem in the field. If
> > anybody did have a wraparound they'd be getting bleats from VACUUM, and no
> > one has reported any such thing that I've heard.
>
> I hit the issue, but Andres helped on IRC (and so started this
> thread). Yes, VACUUM failing on a production server after the 9.1->9.3
> upgrade was the problem. I'd assumed there was database corruption and
> I had some unreadable blocks, and probably would have just rebuilt the
> table if #postgresql hadn't been so helpful.

This is actually a second problem that came up during the investigation
of the problem you had - with different symptoms and causes... IIRC it's
not what you've experienced.

If you look at
http://www.postgresql.org/docs/devel/static/release-9-3-5.html your
problem should be described.

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services