I am hoping to start a small discussion on the feasibility of creating
clusters of slony clusters.

For a future project I am looking at the feasibility of creating a
geographically separated, replicated system of databases.  There would
be a handful of sites, each replicating data that they own, to all other
sites.  This is nothing unusual, and slony seems to work very well for
doing this.

The difficulty arises when I want to use a 2-database slony cluster at
each site to allow for improved reliability, reduced maintenance
windows, etc.  Doing this, I will double the number of databases in the
cluster, and increase the number of interconnections by a huge
(unworkable) factor.

We could considerably reduce this interconnection overhead if we could
consider each site's pair of databases as a single slony node from the
point of view of each other site: to have in effect, a cluster of
clusters.

This would bring a certain assymetry within the nodes of the
supercluster.  Eg, given a supercluster of three clusters as follows:

(A1<->A2) <-> (B1<->B2) <-> (C1<->C2)

Each node would see a different cluster configuration:

A1 and A2 see a cluster consisting of A1, A2, B, and C
B1 and B2 see a cluster consisting of A, B1, B2, and C
C1 and C2 see a cluster consisting of A, B, C1, and C2

It seems to me that extending slony to be able to do this would be
possible, though perhaps not easy.  The bigger questions might be: 
- Is this desirable?
- Will it open a huge can of worms?  
- Are there better ways of achieving the same ends?

All feedback, thoughts, questions of my sanity, etc will be welcomed.

__
Marc
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://gborg.postgresql.org/pipermail/slony1-general/attachments/20061207/b23483d5/attachment.bin