Rebuilding a merge publication & adding a subscriber

Assume the following scenario: I have an existing publication using merge
replication with ServerA as the distributor and publisher and ServerB as a
push subscriber. I need to add ServerC as a push subscriber. In addition, the
publication has been in service for several years, and I understand that it
is desirable to start from scratch by deleting and rebuilding it. My plan is
to create an entirely new publication involving a few minor schema changes
from the existing one, again with ServerA as the distributor and publisher
and with ServerC as the sole subscriber. Once this setup proves functional, I
then plan to delete the original publication and add ServerB as a subscriber
to the new one. Here are my questions.
1. First of all, is this a workable scenario? Are there practical gotchas
that I will run into? Is there any problem with having two publications –
running side-by-side for just a few days – that are nearly identical?
2. Is there any concern that during this interim period of a few days – with
ServerA hosting both publications – the processing load on ServerA would be
excessive?
3. Can the existing distribution database serve for both publications, or is
it necessary or desirable to create a new distribution database for the new
publication?
4. All the publication articles already exist on ServerB and ServerC, but I
want to remove them from these subscribers and push them back out from
ServerA when the new subscription is established, so that a number of minor
schema changes can be included. What is the best way to remove the articles
from the subscriber servers?
Sorry for the monster posting, but I thought it best to get all this out
there at once. TIA...
You cannot publish the same table more than once. But if I understand
correctly, You are going to create a new set of articles that have
slight changes so that should not be a problem. The distributor can
handle multiple publications without any problems and many publications
can run on the same server, so long as the same tables are not
published more than once. The load would mostly be IO and network
traffic but without knowing the kind of data and server details it's
hard to say how it will impact your server. If you have all the data at
the publisher you can drop the subscriptions at the subscribers, then
drop the tables. Script the tables form the publisher and create them
on the subscriber. The merge snapshot will sync the tables and you
should be good.