[parliament32]

I'm having trouble with some of what you're saying. Why is RAID 5 any different than, say, RAID 1? You monitor them both the same way (your raid controller tells you it's degraded), and you fix both the same way (slam in a new drive and rebuild). Why would monitoring/recovery be any different than any other RAID level on a controller?

[dylan604]

RAID 1 is just direct copies of the data onto 2 separate drives. No parity is involved. RAID 5 breaks the data into N-1 segments and then has parity data on the remaing disc. The drive the parity data gets saved to rotates through the drives in the array so that no single drive has all of the parity data. That's what RAID 3 does.

So, if you have a RAID 5 array where 1 drive dies, the rebuild process is meant to be able to recreate the missing data on the new drive using the parity data from the rest of the drives in the array. However, if one of those drives has read errors, typically CRC type errors, then the rebuild cannot continue because the data it needs is not available. Bye bye data!!

Also, RAID 5 can achieve higher throughput since the data is split between the number of drives in the array. We used to build RAID 51 volumes which is 2 separate RAID 5 arrays, but then add the 2 volumes as RAID 1 volume so they are mirrored. The system only sees 1 volume with the capacity of a single RAID 5 volume. That was before RAID 6.