The low-level error will usually be detected first by the disk driver, then it will percolate up to md. So something like:
Code:
kernel: end_request: I/O error, dev sdb, sector 2045799
kernel: raid1: Disk failure on sdb5, disabling device
Operation continuing on 1 devices
Set up mdadm in monitor mode to send a message when an event occurs. I hear all the time from people who set up RAID and their system fails. When you look in the logs, the first drive failed 6 months ago, but they weren't checking.