hannu at skype.net wrote:

>>On 2/23/2005 9:07 AM, hannu at skype.net wrote:
>>    
>>
>>>I wonder if one database be can part of two clusters, if the tables
>>>belonging to these cliusters are kept separate ?
>>>      
>>>
>>Yes, in that case each of the origins runs one slon and the
>>dual-subscriber runs 2 of them.
>>    
>>
>
>Thats what I thought, I was just afraid there may be some hidden
>dependencies.
>
>Thanks!
>
>  
>
The use of namespaces for the Slony-I-specific data is a pretty slick 
way of making it fairly silly for there to be such dependancies.

>>The same effect can be achieved by keeping the tables just in separate
>>set's, each with a different origin. The two origin databases need not
>>be subscribed to each others sets, and the third "consolidation"
>>database can be subscribed to both.
>>
>>It's not 100% the same. If one of the origin databases is down, the
>>other will not clean out old replication log information any more, so
>>the two cluster install has it's pro's.
>>    
>>
>
>It also helps against exponential growth of PATHS's and LISTEN's :)
>  
>
It's merely polynomial :-).