core stack:
root@db4 / $ pstack ~postgres/core
core '/opt/postgres/core' of 19868: pg_upgrade --verbose --link --old-d=
atadir=3D/opt/postgres/db/root/old --
fffffd7ffeda1148 memcpy () + 6b8
000000000040b8b6 transfer_single_new_db () + fa
000000000040b6ea transfer_all_new_dbs () + 116
000000000040ae62 main () + 106
000000000040580c ???????? ()
As to the ownership, the bash script I am testing 9.1.4 and 9.2.0 with recu=
rsively chowns the directory that owns the old and the new PGDATA directory=
before running pg_upgrade.=20=20
Mike Wilson
mfwilson@gmail.com
On Jul 15, 2012, at 2:45 PM, Tom Lane wrote:
> Mike Wilson <mfwilson@gmail.com> writes:
>> I've had some time to examine this closer over the weekend. It
>> appears that pg_upgrade for 9.2b2 segfaults which more than likely has
>> something to do with the resulting converted database appearing to
>> have no rows.
>=20
> Yeah, more than likely :-(. Could we see a stack trace from the
> segfault?
>=20
>> Of possible note in this DB is that the previous DBA renamed the
>> "postgres" user.
>=20
> Hmm. There is a known bug in beta2 that's triggered by old and new
> clusters not having the same name for the bootstrap superuser; although
> I don't recall that the symptoms included a segfault. In any case,
> I'd suggest making sure the new cluster is initdb'd under the same
> account that currently owns the old cluster.
>=20
> regards, tom lane
