If the data is read-only, I use lsyncd to do it and have the data on the local disk. It is much cleaner, faster, depends only on ssh and gives you a backup as the source data is not on either droplet. It also means lower cost droplets.

If some areas of the data are writeable, I use NFS or else unison. I tried GlusterFS, ceph and DRDB and they all have drawbacks, the biggest of which is the need for a PhD in Arcane Computing to get them to work.