On Thu, 29 Jan 2009, Bruce Momjian wrote:
> One thing I think would help would be a pg_class column that says
> whether the table is ready for upgrading.
Some quick reminders here; Tom included that idea from one of your earlier
discussions but also suggested a similar column in pg_database:
http://archives.postgresql.org/pgsql-hackers/2008-11/msg00436.php
The main objection I saw was Heikki suggesting that this isn't necessarily
forward progress if that's not actually sufficient to save all the state
information needed:
http://archives.postgresql.org/pgsql-hackers/2008-11/msg00449.php
This particular idea keeps coming up though, and I would imagine the worst
case here is that it gets included in 8.4 but isn't enough; it's gotta at
least help open options for future in-place upgrade work.
(The other complaints revolved around multi-version upgrades, which are a
pipe dream now anyway)
Zdenek suggested that "latest processed block" would also be useful here:
http://archives.postgresql.org/pgsql-hackers/2008-11/msg00540.php
P.S. If anybody would like a 275 message "best of" archive just following
the last four months of the pg_upgrade saga, let me know off-list; that's
how I found all those so fast. As just that subset consists of 22
threads, it's kind of messy to sort through most other ways.
--
* Greg Smith gsmith@gregsmith.com http://www.gregsmith.com Baltimore, MD