My suggestion would be to upgrade to 9.1, then use the internal ADP drives to create a single aggregate with a single RAID-TEC raidgroup of 23 disks (with 1 spare, or 24 without) for ~95TB usable, and then on the other controller, use the external drives in the same sort of configuration - single aggr single RAID-TEC raidgroup with ~96TB usable.
I think this is pretty much what I want to do with a freshly rebuilt FAS2240-4 on ONTAP 9.1 with 48 x 2.42TB Nearline drives.
All disks are the same. All odd number disks are owned by node 1, evens by node 2. We have root-data partitioning of the 24 disks in shelf 1, and the 24 disks in shelf 2 are unpartitioned spares. Both root aggregates are apparently 95% full with 180/189 GB used, and they each have 2 spare slices of 27.72 GB.
It seems from reading http://docs.netapp.com/ontap-9/topic/com.netapp.doc.dot-cm-psmg/GUID-07302AD3-F820-48F7-BD27-68DB0C2C49B5.html
that I can reassign all the Data partitions of shelf-1 to node 1 (leaving the Container and Root partitions owned by alternating nodes) and reassign the Container partitions in shelf 2 to node 2. Then I'll be able to make the wide-striped 20+3P raid groups that you suggest, with the shelf 2 Data slightly bigger than shelf 1.
1. Have I read that correctly?
2. Is there any downside to having different nodes own the Container, Root and Data partitions of some disks?
3. Should I be worried about the 95% full root partitions? This is freshly built and automatically partitioned. Do they need to be expanded?