Sql-server – Enable Replication/CDC from High Availability Group

The link posted as a comment to your question explains the technical steps but may not have made clear how the setup really works.

When you create the subscription you inform the publisher that the subscriber has been restored from a backup (specifying the backup file used). The backup you use has to be taken after the publication was created. When you setup replication you tell the distributor how long to keep transactions around (Transaction Retention Period). The distributor then keeps these commands until they are delivered to all subscribers, or until the retention period has been reached. So when the subscription is created, the LSN of the backup is checked to see if it was taken from within the retention period. If so, then the backup and therefore the subscriber is valid. Only the undistributed transactions created after the backup will need to be sent to the subscriber.

In your case you would:

• setup the publication
• take a backup
• copy that to the subscriber before the retention period expires
• restore the backup to the subscriber
• create the subscription (using the initialize from backup option)
• watch as the undistributed commands then get sent to the subscriber
• finally see that the entire thing works!

Initializing from backup makes the entire thing much easier than snapshots, because you generally have backups running anyhow and, more often than not, can use compression. This does mean that the entire DB is on the subscriber though, which may be desired or may not.

Put the distributor in a virtual machine, and use the hypervisor's high availability tools. This lets you survive host hardware failure. OS problems will still cause issues, though, so for an additional level of protection, you can run clusters inside virtualization.