Re: [Gnumed-devel] database replication with bucardo

[Busser, Jim]

On 2012-06-27, at 1:10 PM, Karsten Hilbert wrote:

>> Bucardo does a trigger based replication, so is nearly on time.
> 
> Regardless of whether it's on time - it will be serialized
> meaning that every single row change will be transported
> which means there should be no need to explicitely replicate
> audit tables.

Um... I can see how new row insertions will avoid any problem as the audit 
tables will not have been touched.

Does the following properly describe how audit tables will remain in order?

Under Bucardo,

- each copy of the database will act as a "replicate" of the other, permitting 
changes at either end to be reflected in the other (IOW, the model is not 
master-slave but master-master)?

- the audit tables need not be included in the synchronization, because in each 
case where:

        a row gets modified at one end, AND
        a row change will be mirrored in the replicate

the local audit table will get updated as a function of the local row change?

-- Jim