[mythtv-users] Why RAID 5 stops working in 2009

[Yan Seiner]

Ian Clark wrote:
>
>
> 2008/10/23 Another Sillyname <anothersname at googlemail.com 
> <mailto:anothersname at googlemail.com>>
>
>     2008/10/23 Mache Creeger <mache at creeger.com
>     <mailto:mache at creeger.com>>:
>     > From Slashdot
>     >
>     [snip]
>
>     >
>     >
>
>     Given the amount of time it would take even a hardware controller to
>     reconstruct a 2TB drive you'd be unwise to use them.
>
>     750GB drives using a decent hardware controller take about 30 hours to
>     reconstruct a failed drive under raid 6.....this means you have a
>     'failure window' of those 30 hours when a 2nd failure kills you.  I
>     can't imagine I'd even want this 'failure window' to be in the 72+
>     hour timeframe.
>
>
> Erm, if it's raid 6, wouldn't you need 3 drives to fail for it to kill 
> you?
The way I understand RAID-6, the failure would have to be:

1 dead drive
1 URE on another drive
1 URE on yet another drive in the same stripe as the above URE

So the window is fairly small as it would require specific failures on 
specific parts of 2 drives.  Yes there is a risk, but if the data is so 
valuable, then make a backup, mirror it to another array, or hire me to 
back it up off-site for you.  :-)

The question really becomes, is there some way to stripe the array to 
minimize the risk of dual failure in the same stripe?  What are the 
performance/reliability hits?

--Yan