Reminder -- RAID 5 is not your friend

[richard.pieri-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org (Richard Pieri)]

On Mar 10, 2010, at 11:27 PM, Dan Ritter wrote:
> 
> Before we can get around to it, another disk in the storage
> system also dies. Poof.
> 
> We have replaced that entire group now with a RAID 10. We no
> longer have any RAID 5 setups.

You've thrown money at the problem without improving reliability.  RAID 10 is not more reliable than RAID 5.  RAID 10 requires a minimum of 4 disk.  A1 is mirrored to B1 and A2 is mirrored to B2.  *1 and *2 are then striped together.  What happens when you lose A1 and B1?  Answer: you lose your data.

The benefit to RAID 10 (RAID 1+0) is performance.  It's faster than RAID 5.  It costs more because you need N*2 disks instead of N+1.  Both provide the same level of reliable fault tolerance: a one disk failure.  There are nested RAID levels that provide greater fault tolerance but no matter how much redundancy you have a catastrophe will take you out.

That's the lesson you need to take from this and drill into your management.  Catastrophes happen.  Plan accordingly.  That includes time to restore from backups.

--Rich P.