In HNAS, we have files based and object based replications. Object based replication is much faster and it can replicate only the full filesystems and not the individual shares or directories. And also there are some limitations when it comes to data migrator and deduplication features. Even though the file based is bit slower than object based, it is very much compatible with data migrator and deduplication features.
With object based replication, we can swap the roles by promoting and demoting the file system versions based on the availability of NDMP snapshots. This is very easy and we can easily reverse the replication by replicating only the changes happened on the new target.
But with file based replication, I don't see any easiest method to swap the roles and replicate only the changes from new primary to old primary. I think there is a script called "replication_recovery" available on SMU to achieve this requirement. Can someone explain this.