The problem with this solution is that you cannot use both databases at the same time and that you are not sure if all data has been totally replicated to the second node (or you need XA-style replication).
In this case, you could also imagine a single partition with a standy-by database. This can be achieved by using SAPDB (for example) open source database while keeping a safe version of the journal for example.
In any case, take a look at the DB solutions proposed by Interbase/Firebird, PostresSQL and SAPDB (sapdb.org).
> In this case, you could also imagine a single partition with a standy-by database
Actually, if I got the picture right, what Dany proposes is something almost like that: The second partition is a stand-by replica of the first one.
Dany, what is the point of having a stand-by replica of all environment? Particulary on a budget environment, wouldn't it be better to use all app-server power you could? Why have 2 app-servers idle until something fails in the first partition?
Just my tough about this...
I actually think that is waist of $$ wathever kind of passive replication (where "passive" means replicas that are not used for whatever purphose unless something breaks)
Yes, ideally you would use all the server power you have available and with a clustered database that could be achieved. If you are using an open source database however, you are stuck with replicating the database to keep from having the database be a single point of failure. If you know of an open source or free database that clusters or has something better than replication, please describe. Having 2 partitions with 2 nodes and each partition with 1 database is an example that ecompasses load balancing and fail over using database replication. In this example configuration, if you could purchase a clustered database for less than $2,500 for 2 server licenses, then that would be the way to go.