failure (specifically bad sectors) and replace the drives when more than a few of these happen.
The three biggest problems imho with RAID are:
- 1. The controller issue you mention above -- we use software RAID on linux for this reason (most of the time)
- 2. Do you know when a drive is actually dead? It can be mostly dead and you don't know unless it gets kicked out of the array. It's scary how often 2 or more drives get booted from an array in short succession too. SMART monitoring is a must (Google did a major study on HD risk failures years back - *any* bad sectors at all, and particularly rate of change in bad sectors were the single biggest indicators of near-certain drive death within 6 months)
- 3. What happens (specifically to the various layers of caching) under power loss
#3 in particular is why some spend big bucks on expensive RAID solutions, #1 to an extent as well and particularly if you want 'performance'. You don't need this level of crazy, but you still need to be aware where your risks are so you can work around them.