I have a 3 node cluster running on Windows Server 2012 R2. Two nodes for the databases, the 3rd node is an Always On db used for reporting. I have new hardware built running the same OS, but will need to upgrade the SQL Server version from 2012 to 2016. All servers will utilize the same SAN.
Should I stand up a new cluster running SQL Server 2016, restore the database to it, going through all the renaming and such? Or can I do a type of rolling upgrade where I join the new servers to the existing cluster (I guess I would need to install 2012 on them first and then kick off the upgrade to 2016 on them) once I fail over to the new servers, I would evict the old nodes without upgrading them?
This seems like it would have the least amount of downtime. But I'm not sure if I can easily rejoin the old servers to the cluster and rollback to the 2012 instance if there are problems with the 2016 instance. Most whitepapers I see revolve around OS changes, not same OS, but new SQL versions.