> You have recreated what pg_upgrade does. It is for upgrading system
> tables. If only your system tables were hosed, you are fine now.
Only if your previous system has been vacuum'ed and no dml afterwards.
Otherwise you also need to copy your old pg_log.
>
>
> > The Hermit Hacker wrote:
> > > just a quick thought ... have you tried shutting down and
> restrating the
> > > postmaster? basically, "reset" the shared memory? v7.x handles
> > > corruptions like that alot cleaner, but previous versions
> caused odd
> > > results if shared memory got corrupted ...
> >
> > Well, I've rebooted twice. In fact, it was a hard lock that
> caused the
> > problems. When the machine was brought back up, the db was foobar.
> >
> > I'm doing something really really evil to avoid losing the
> last days'
> > data:
> >
> > -I created a new db
> > -used the old db schema to create all new blank tables
vacuum new db
(I would do a tar backup of the whole old db)
vacuum old db, if that is possible
> > -copied the physical table files from the old data
> directory into the
> > new database directory
if above vacuum old db was not possible copy old pg_log
> > -currently vacuuming the new db - nothing is barfing yet
> > -now hopefully I can create my indexes and be back in business
> >
> > Tim
Andreas
