I have to replicate over 50GB of data over a slow network. I did not use the option to initialize snapshot from database backup because the replication articles contain row filters. If I do that, I'll have to run a lot of scripts to remove the
data and other unnecessary database objects on the subscriber.
Instead, I created a workaround. On the publisher, I first create the actual push subscription to the target subscriber on the publication. This subscription is set not to initialize from snapshot. I then created a second push subscription
on the same publication, but the subscriber is a random database on the publisher server. This second subscription is a dummy subscription set to initialize from snapshot - the purpose is to generate the necessary snapshot files. I then reinitialized
all subscriptions and generate the new snapshot files.
On the subscriber, an empty database is created with the same tables as the publisher database. I created an identical publication on this empty database, and a dummy push subscription on the target subscriber. The subscription is reinitialized,
and the snapshot files on the empty database is created. These dummy snapshot files are then overwritten with the actual snapshot files created on the publisher, and then I synchronize the the dummy subscriptions with the actual sn
View Complete Post