Andrew -

Thanks for pointing out the subtly.  I may well have violated that....

John

Andrew Hammond wrote:
> On 3/8/07, Darcy Buskermolen <darcyb at commandprompt.com> wrote:
>> On March 8, 2007 05:31 am, John Goetsch wrote:
>> > The question is:  should all the nodes be one cluster ie master, slave
>> > and all subnodes, or should I have one cluster for the master-slave
>> > replication and one cluster for the slave-multinode replication.
>> >
>> > I can't seem to find the policy/rules/conventions about cluster 
>> creation.
>>
>> A single cluster with multiple sets.
>>
>> One set is the subset of relations that you want to cascade, the 
>> other set is
>> the non cascaded relations.
>
> As Darcy says, however be aware that you need to have the exact same
> schema on all the members of the cluster. For example, you need tables
> to exist on the slaves even if you're not replicating data into them.
>
> See the email to slony1-general list on 2007-Feb-20 from Sven
> Willenberger, subject "EXECUTE SCRIPT operates on non set member
> tables". I'd link to the mailing list archives, but they seem to be
> broken following the move.
>
> Andrew
>