Re: Notes on testing Postgres 10b1

Hi,

On 07/06/17 07:01, Josh Berkus wrote:
> Folks,
>
> I've put together some demos on PostgreSQL 10beta1. Here's a few
> feedback notes based on my experience with it.
> [...snip...]
>
> Problems
> --------
>
> P1. On the publishing node, logical replication relies on the *implied*
> correspondence of the application_name and the replication_slot both
> being named the same as the publication in order to associate a
> particular publication with a particular replication connection.
> However, there's absolutely nothing preventing me from also creating a
> binary replication connection by the same name It really seems like we
> need a field in pg_stat_replication or pg_replication_slots which lists
> the publication.
>

What do you mean implied correspondence of application_name and the
replication_slot? We only use subscription_name as default value for
those when user does not specify something else, all three of those can
have different value if user sets it up that way. And there is no
correspondence whatsoever to names of publications. The upstream only
knows which publications to replicate because subscription gives list of
requested publications as option to START_REPLICATION walsender command.
The list of publications associated with a subscription are only stored
on the subscriber and publisher has no idea what those are.

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