Thanks for your answer. Unfortunately the tables do not have primary keys
(not my design!) so I can't use transactional replication. I believe DTS
"Hilary Cotter" wrote:
> when the snapshot is applied on the subscriber it truncates the data in the
> table. There is no way to rollback to what was there before.
>
> I think you should use transactional replication as only the changes are
> replicated.
>
> DTS is not transaction aware, so it is all or nothing. DTS will be faster
> when you are sending a snapshot to a single subscriber. For more than one
> subscriber, snapshot replication will be faster.
>
> --
> Hilary Cotter
> Looking for a SQL Server replication book?
>
http://www.nwsu.com/0974973602.html > "Learner" <Learner@discussions.microsoft.com> wrote in message
> news:7A8CAAA5-6ABB-470D-8318-3A2879684E87@microsoft.com...
> > I'm learning replication and am wanting to check if my understanding is
> > correct:
> > If snapshot replication is set up with 'delete all data' for article name
> > conflicts and the replication fails midway eg from connection failure
> > (after
> > existing data is deleted and before new data is inserted), could the
> > subscriber be left with an empty table or would it automatically rollback
> > to
> > the previous data?
> > The reason I ask is that I need to copy data from a couple of tables to
> > several databases every night. I will not know if the data has changed
> > from
> > night to night (but is only few hundred rows anyway) so will have to copy
> > all
> > the rows and can live with old data being present but NOT no data being
> > present.
> > Is snapshot replication the right way to go? I was also wondering about
> > creating a DTS package that I could use transactions in?
> >
> > Thanks for any help
>
>