[KLUG Members] Re: RAID

[Mike Williams]

>
>
>From: Adam Williams <awilliam@whitemice.org>
>To: members@kalamazoolinux.org
>Organization: 
>Date: 14 Nov 2002 22:00:49 -0500
>Reply-To: members@kalamazoolinux.org
>
>  
>
>>> We have a 1TB RAID5 array that lost two drives at the same time. 
>>> <sigh>
>>    
>>
>
>Yikes, very bad karma.
>
Bad luck, yes, but it's far from unheard of.  In RAID arrays with all 
new drives, having several of them fail in close proximity is pretty 
common.  The drives are the same make and model, usually, and sometimes 
even consecutive serial numbers, or nearly so.  In RAID 5 they get 
identical usage, so with tight manufacturing specs (which hard drives 
must have) it's reasonable that identical drives under identical loads 
will fail at pretty similar times.

>>> It was on our critical file server. Needless to say panic followed and 
>>> schedules were delayed....
>>> We bought an identical unit and a different brand of drives just for 
>>> diveristy. We really wanted to now mirror two RAID5 arrays.
>>    
>>
>
>Thats RAID 51 (mirrored RAID 5).  Why not just RAID 50?
>
You mean besides the fact that 50 would give you a higher failure rate 
than 5?  If you stripe a pair of RAID 5's, you only need to lose one of 
the RAID 5's to take down the array.  Striping worries me with todays 
drives that can fail without warning when they fail.

>....
>  
>
>>> So this brings me down to running one RAID as the active drive and 
>>> mirroring the changes each night to a second identical RAID drive.
>>> Is there a program for this? Is there a recommend way to keep two drives 
>>> in sync? I thinking a cron job?
>>    
>>
>
>If your distro has LVM support this should be pretty easy.  Snapshot &
>backup/restore to the other array.... but why not just use tape?  If it
>isn't "hot" this is a pretty serious solution.
>

How about RAID 5 with a hot spare?  That will protect you from 
sequential failures, and it will be less costly than (And, I think, less 
messy) than trying to software mirror two arrays.