First of all, the quality of the md kernel driver (the component that provides RAID fucntionality) is exceptional, and the data rates that i was getting when testing this setup rivalled those of HW controllers quite favourably...
Ok, now that this is out of the way, *IF* one of your disks fails, you will not have to do anything. md will flag the disk as failed (you'll be able to see this by doing cat /proc/mdstat) and when you replace the disk, md will pick this up and rebuild it as a mirror (you'll also be able to see the resyncing process by doing cat /proc/mdstat).
As a sidenote, if you really want to get into admin of RAID servers, you should get the new, updated tools for using RAID via md. They are called mdadm and they have a plethora of options to do just about anything with RAIDs.
The url :
http://www.cse.unsw.edu.au/~neilb/source/mdadm/
Here you can get the tarballs, and (maybe) rpms too. The old (i.e. 'mkraid') tools are not supported anymore and if you want to use RAID you should use mdadm.
I hope this helps,
Cheers,
Vlad.