It is surprising to see the number of people that implement some sort of redundancy with no means of measuring when one of the components has failed. For example, building a disk mirror, but never knowing when one of them has failed.
That’s one thing that’s kept me away from using Linux Software RAID in anything serious, since I never saw a good way to determine if anything had failed until I happen to stumble upon an odd log entry that ends up meaning a failed disk.
Sys Admin magazine has a good article on software RAID where they actually tell you how to monitor for a failed disk and how to rebuild it.
I should also note that Fedora includes smartmontools, which monitor SMART capable drives (and most are) for both faults and prefaults. It somewhat makes the former link a moot point, but since you’re building redundant disks, you may as well have redundant ways of checking for failures.