Replication - should destination server snapshot or not?

[pvuchetich]

Does anyone have specific guidance on whether or not snapshots should be run on a dataset that is the destination of replication?

In the current (9.3) documentation, for section 8.3 "Replication", there is not guidance on whether or not to run local snapshots on the "PULL" (destination) server for snapshots.

With my initial testing of this functionality, I have sent it to a new pool that is *not* running snapshots because I am not certain if using two sources of snapshots ("PUSH" server and "PULL" server) might confuse the "PUSH" server replication process and/or the "PULL" server snapshot cleanup process, especially if the snapshots might be named with the same timestamp, or if it might cause other problems.

 

[cyberjock]

I would definitely not "double snapshot" datasets you are replicating. I'm betting the replicator could figure it out, but it's still just "weird" in my book.

 

[pvuchetich]

Thanks,
The "double snapshot" is what I was concerned about. Even if it didn't cause errors, I might get confused about which snapshot is which when trying to recover using the remote backup.