Re: Logical replication existing data copy

On 20/03/17 21:25, Stephen Frost wrote:
> Peter,
>
> * Peter Eisentraut (peter(dot)eisentraut(at)2ndquadrant(dot)com) wrote:
>> The current patch causes a failure in the pg_dump tests, because the
>> generated CREATE SUBSCRIPTION commands make connection attempts that
>> don't work. We have the pg_dump option --no-create-subscription-slots
>> for this, but I suppose we should expand that to
>> --no-subscription-connect and use the new NOCONNECT option instead.
>
> I'll admit that I've not followed this very closely, so perhaps I'm
> misunderstanding, but I thought our prior discussion lead to the idea
> that pg_dump would always dump out subscriptions with 'NOCONNECT', so
> that a restore wouldn't automatically start trying to connect to another
> system.
>
> In which case, I don't quite understand why we need a
> "--no-subscription-connect" option.
>

Yeah I think that was the agreement as well.

Here is fixed version, also rebased on top of all the changes to pg_dump
tests. Subscriptions are dumped unless --no-subscriptions is specified.
They are dumped with NOCONNECT option (which didn't exist before this
patch so it was bit harder to do the dump by default).

--
Petr Jelinek http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services