I would appreciate any comments and advice on the follwoing scenario.
We have a number of databases with data being inserted into a small number of tables at a fairly high rate. We want to use transactional replication to send the data to a central database, merging the data into an identical table structure. Where required, IDs are already being managed by assignning a generous range of IDs to each field database. The connection to the central database is via a WAN and can be interrupted from time to time.
I have a number of questions:-
1. If we have the Log Reader agents running on field database servers (i.e. the publisher), what happens to transactions if the WAN is down? Do they simlply accumulate in the transaction log or does the Log Reader agent process then into replication commands and store then in a local file or table?
CORRECTION: Actually I don't think that the Log Reader can run on the publisher, only on the distributor or the subscriber (which will be one and the same in this case). So the question reamains, what happens when the WAN is down? I assume that the transaction log will grow until the Log Reader eventually clears it out. Is one option to keep the transaction log smaller to have a distributor running at each publisher? It sounds as though this would be more difficult to set up, maintain and monitor.
2. Even though this is transactional replication, we will be actually merging data from the publishers at the subscriber, which means we NEVER was a a snapshot to be replicated. Is there a way to completely suppress snapshot replication? It seems to be enabled by default.
Any other words of advice would be appreciated.
R Campbell