DR Testing

[Kevin Horton]

After all my fun using the CLI to do replication when I moved data from one server to another, I now use the GUI exclusively. Create a periodic snapshot that recursively includes all the datasets of interest. Then follow the docs to use the "Semi-Automatic" setup to replicate the data to your other machine. Once all the replication has completed, you can disable the replication task by unchecking the "Enabled" box in the Replication Task.

 

[Derek Humes]

This is what I had in mind for an offsite backup, but leaving replication checked so that daily my snapshots will get sent offsite. I just didn't know that the deletion of stale snapshots could backfire in a way like this during a disaster recovery scenario - although admittedly I don't know for certain that is what caused my failure.

 

[Kevin Horton]

I too am completely baffled as to what happened. But, like you mentioned, this is why it is important to test the backup recovery.

Having more than one backup is also a very good idea. My new main server has two pools, with the second pool being a backup of the main pool, with hourly replications. The old server is currently being tested as a potential off-site backup server, with nightly replications.