Hard Disk IDs in Linux

Lonnie Olson lists at kittypee.com
Mon Mar 11 09:58:35 MDT 2013

On Mon, Mar 11, 2013 at 5:35 AM, Dan Egli <ddavidegli at gmail.com> wrote:
> The idea of stringing the RAIDs together has it's merit. Would I just mdadm
> to create a separate /dev/mdX out of the various mdXs already created? I.e.
> suppose I put eight drives per array. Using a 32 disk rackmount case gives
> me a result of md0 - md3. Would I create an md4 that's a raid 0 of md0 -
> md3? Or would I perhaps be better off using logical volume manager (which
> I've never used before but can look up if needed).

Either of those two options would work fine.

> The thing I was trying to accomplish with the larger array was the large
> amount of disk space available. I.e. if I fill the enclosure with 32 4TB
> drives and make a single raid 6 I wind up with 120TB of available storage.
> On the other hand, if I was to make four equal array of 8 drives each in
> raid 6 then I loose not just 8TB total but 8TB per array, or 32TB total,
> leaving me with 96TB, which is still a good amount of storage, but also
> quite literally only 75% of the disk capacity. If anyone knows a good way
> to increase the disk capacity availability to raise it above 75%
> (preferably back up to the 90% range) I'm all ears.

True, you do lose capacity by splitting into smaller arrays, but it
gives you tons more protection from failure.
With 1 array, you can only lose a maximum of 2 drives before the whole
thing fails.  With 4 arrays of 8 drives, you can lose a lot more
depending on which ones fail.

It's a risk trade off, capacity vs protection.  How important is the
data?  How much money would your org lose in the time it took to
restore all the data from backups.

There are other middle options depending on where you want your
capacity/protection line to be.
* 2 RAID6 of 15 disks, with 2 hot spares (1 each) - 104TB
* 1 RAID6 of 30 disks with 2 hot spares - 112TB
* 1 RAID6 of 32 disks with 4 cold spares, perfect RAID monitoring that
paged you instantly, perfect response w/in 4 hours to be onsite to
replace drive - 120TB

To me, a few TB lost to protection is worth a lot.

More information about the PLUG mailing list