On Thu, Aug 9, 2012 at 09:20:23AM -0400, Robert Haas wrote:
> On Wed, Aug 8, 2012 at 7:04 PM, Bruce Momjian <bruce@momjian.us> wrote:
> >> The point I think Robert was trying to make is that we need to cut down
> >> not only the complexity of running pg_upgrade, but the number of failure
> >> modes. At least that's how I'd define improvement here.
> >
> > Agreed. Even with these changes, I still see a lot of complexity.
>
> I agree. That's why I said it needs some serious engineering time to
> file down the rough edges, plural, not that it needs this fix in
> particular. This would help to make things less error-prone, but it's
> far from the only thing that is needed. As to what exactly is needed,
> well that's up for discussion.
>
> One of the big failure modes for pg_upgrade is... pg_dump's dump fails
> to restore. That bothers me quite a bit because there are actually a
> lot more people who rely on pg_dump than there are people who rely on
> pg_upgrade, and it turns out there are all of these edge cases that
> pg_dump doesn't actually handle all that well. Sure, you can edit the
> dump by hand (if you're not using pg_upgrade) but that sucks.
100% agree. 9.2 will improve pg_upgrade debugging capabilities and
speed, but there are still many rough edges where that I don't see clear
solutions. Therefore, I don't see pg_upgrade being simple anytime soon,
and hence, it is also unlikely pg_upgrade will be moved to /bin anytime
soon, if we use the same criteria.
-- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB
http://enterprisedb.com
+ It's impossible for everything to be true. +