> init cluster ( id=1, comment = 'Master Node');
>
> table add key (node id = 1, fully qualified name =
> 'cagold.address_history');
>
>
> create set ( id=1, origin=1, comment = 'PA REPLICATION SET' );
>
> set add table ( set id=1, origin=1, id=1, fully qualified name =
> 'cagold.address');
> set add table ( set id=1, origin=1, id=2, fully qualified name =
> 'cagold.address_history');

> 5) I reset everything, and re-run and I get the same error:
>
> <stdin>:15: PGRES_FATAL_ERROR select
> "_ca_cluster".determineIdxnameUnique('cagold.address_history', NULL);
> - ERROR:  Slony-I: table "cagold"."address_history" has no primary key
>
> to "reset" I did the following:
>
> a) dropped (with cascade) the underscore schema named after the
> replication cluster in the master
> b) dropped the slave db
> c) re-setup the slave
> d) re-ran the above script
>
> Any thoughts ?

FYI, there was no need to go back to the very beginning in this process
(e.g. - to drop replication and drop databases and such).

You did correctly notice that you didn't have a primary key; as observed
elsewhere, the *RIGHT* answer is to add one yourself.

TABLE ADD KEY is considered something of a misfeature, and has been
removed in v2.

But if you feel like you really MUST use TABLE ADD KEY, then you can do
so; you should look at the documentation for SET ADD TABLE, as you'll need
to modify the command to indicate that you're using TABLE ADD KEY on the
table...

http://linuxfinances.info/info/stmtsetaddtable.html

Note that once you submit a TABLE ADD KEY request, this is NOT an
idempotent operation, and this will *break* things if you try to "clean
up" via dropping the replication schema.  Dropping the schema won't make
the slony-created key column go away.