Re: pg_upgrade ?deficiency

> Bruce Momjian <bruce@momjian.us> writes:
> > Not sure about backpatching.  default_transaction_read_only has been
> > around since 7.4.  Setting it to true would cause pg_dump to fail unless
> > you changed the database setting, and pg_dumpall would fail completely
> > as there is no way to turn off the database setting.
>
> No, neither pg_dump nor pg_dumpall would fail.  What would fail is
> restoring into a database that has this option already set.  It's possible
> that users of this option haven't noticed it because they never attempted
> a restore in such a context.

I was the original poster on -users who raised this issue. Maybe I can
clarify somewhat:

I have been attempting to upgrade an 8.4 cluster to 9.1
by means of the 9.1 pg_upgrade command.

That failed due to one of the databases in the 8.4 cluster
being "ALTER DATABASE ... SET DEFAULT_TRANSACTION_READ_ONLY TO ON".

Hence my question on that list whether that was to be considered
a bug, a deficiency, or an oversight.

I knew workarounds quite well but wondered whether that
pg_upgrade behaviour was intended to stay that way.

I suggested that if it is intended to stay it might benefit
from a hint in the documentation.

> Yeah, it's a minor issue at best, but perhaps worth fixing since
> the solution is so easy.

That would be really helpful.

> The bigger picture here is that there are lots of ways to break
> pg_upgrade via not-sane settings, and there always will be.

Would setting default_transaction_read_only to on be considered
non-sane ?   If so, why ?

> I don't think we should try to promise that there won't be.

That last assertion is what everyone should certainly be able
to agree with ;-)

Thanks,
Karsten