Devrim GUNDUZ wrote:
> Hello,
>
> Has anyone tried Slony-I on a partitioned table?
>
> I'm working on some tests per a question from a local user. He says that
> he cannot add child tables to replication set, because child the table
> does not have primary keys. Because of DO INSTEAD, no data is inserted
> to master table, so there is no replication :(
>
> Any solution on this?
>
> I thought of adding a trivial serial column to each child table, which
> will act as a primary key;  so that I can add them into replication
> set. 
>
> Does it sound good?
>   
I recently added an example involving inheritance to the set of
regression tests, and it works fine :-).

But note that I expressly added primary keys to each 'sub table':

------------------------------------------------------------------
create table master (
   id serial primary key,
   trans_on timestamptz default 'now()',
   data text
);

create table sub1 (
) inherits (master);
alter table sub1 add primary key(id);

create table sub2 (
) inherits (master);
alter table sub2 add primary key(id);

create table sub3 (
) inherits (master);
alter table sub3 add primary key(id);
------------------------------------------------------------------

The example doesn't add constraints and "DO INSTEAD" triggers; that
would be an interesting addition.