Re: Testing Spare Drives in Software RAID

On Wed, Aug 22, 2007 at 11:05:17PM -0400, Hal Vaughan wrote:
> I know mdadm offers a "--test" command, but it seems quite useless:
>
> -t, --test
> Generate a TestMessage alert for every array found at
> startup. This alert gets mailed and passed to the alert program. This
> can be used for testing that alert message to get through
> successfully.
>
> That helps with knowing that alerts are working, but I'd like to be able
> to test the spare drive in an array. Is there any way to test a spare
> device in a RAID to make sure it's in good shape and not likely to fail
> once it's needed?
>

I don't know how to test the actual functioning of a spare drive. With no activity on a spare, S.M.A.R.T. may not have any data to go on.

You could have an extra spare drive. This spare drive could be as large as your largest partition used in your arrays. You could test this drive off-line with, for example, badblocks to write and read every block of the drive, giving S.M.A.R.T. something to go by. Then you could partition this drive as needed to create a temporary spare partition to put into an array, allowing you to remove a spare drive from an array for similar testing.

If the array is raid1, you could add the spare in so that there are 3 mirror images. Once it has synced, S.M.A.R.T. should have some good data. You can then remove it and turn it back into a spare. I don't know how to do this with raid levels other than 1.

Another method that doesn't require a spare-spare but with some risk would be to force-fail one of the active drives so that the spare comes into play, then add it back in as a spare. The risk is that if a drive in the array actually dies while the spare is being sync'ed then the array has two failed drives and the array itself fails.

Doug.

-- 
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org 
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org