[Pgcluster-general] Table requirements

[Rick Vernam]

On Thursday 17 May 2007 11:22:40 am Gábriel Ákos wrote:
> Hi,
>
> I'd like (for documentational purposes) have an exact statement about
> what a table has to look like to be replicated properly by pgcluster?
what?  I've never heard of such a thing.  pgcluster doesn't bother itself with 
table structures, indexes...etc.  it just executes queries on a group of 
servers over standard psql.

> As far as I learned some kind of unique key has to be there.
I'm not claiming to be an expert...where did you learn this?

> Can this be a simple "composite key" of - say - the columns of the
> table? For example I have a table with 2 string columns (name, role).
> This is not replicated correctly by default. As far as I learned by
> generating a composite key (name+role) it won't work either. By adding
> an (otherwise meaningless) unique id, generated by a sequence, the
> table will be replicated correctly.
>
> Thanks for your help!