Re: pg_group_name_index corrupt?

"Hiroshi Inoue" <Inoue@tpf.co.jp> writes:
> If I remember correctly,pg_upgrade doesn't shutdown the postmaster
> after(or before) moving OLD data to the target dir though it tells us
> the message "You must stop/start the postmaster ...".
> How about calling pg_ctl from pg_upgrade to stop the postmaster ?

What I would actually like to see happen is that pg_upgrade uses a
standalone backend, with no postmaster running at all for the entire
procedure.  Having a live postmaster connected to the system just
opens the door to getting screwed up by some other user connecting to
the database.  But that's a bigger change than I dare try to make right
now...
        regards, tom lane