If you really want a setup with that many disks, you might look into Ceph. It’s intended for handling stupidly huge amounts of data spread across multiple servers with self-healing and other nice features. (As the name suggests it’s a bit of a tentacle monster though.) One of my colleagues set up a deployment at work. It took a while for him to figure out how to get it running well but it’s been pretty useful.

Additional relevant discussion on HN: https://news.ycombinator.com/item?id=39865810