OpenBSD supports root-on-RAID, and has for years, in two ways:
- Hardware RAID. These appear as standard bus devices to the OS. If these hardware devices can register with the bio(4) driver, they can also be managed via bioctl(8). The bio(4) man page lists all of the currently supported hardware RAID technology that can be managed with bioctl.
- Software RAID via RAIDframe. One cannot -boot- from a RAID set, but one can certainly configure -root- on RAID.
There probably isn't a "best" way; but the easiest would be a hardware RAID, I would think.
I use RAIDframe with root-on-raid with one of my systems, and have had many disk drive problems, but never an outage due to them.
I set that RAIDframe server up so it notifies me via e-mail when a failure occurs. I've been e-mailed many times. Each time, I have been able to recover with the same hardware (forcing bad blocks to reallocate before returned the failed drive to the RAID set) or if necessary swap drives, and all without data loss. I have to take a scheduled outage to swap drives, though, as they are not hot swappable.
I have also used raidctl(8) to intentionally break the RAID 1 mirror to manage risky software upgrades/reconfigurations, as recovery is faster than doing a complete restore.
(RAID sets are not backups. But you already know this.)