Preventing I/O starvation on MD1000s triggered by a failed disk.

Stroller stroller at
Tue Aug 24 16:25:43 CDT 2010

On 24 Aug 2010, at 12:54, William Warren wrote:
> ...
> Looks like a classic URE to me.  Modern SATA drives have a read  
> failure
> rate of about 1 per every 12 TB.
> More and more likely with these huge capacity Disks RAID 5 is going to
> hit a URE upon rebuilding which means it fails and the whole array
> dies.  RAID 6 right now is the bet option to get around it.   
> Consistency
> checks aren't going to get around this as you have to read back every
> disk in the array to rebuild.

Can't believe you'd take that Harris bloke seriously. His "RAID5 stops  
working in 2009" article was alarmist enough to get plenty of  
coverage, but it was debunked pretty thoroughly at the time.

Are you seriously telling me that if I go out and buy a 2TB external  
drive from PC World, fill it up with movies, it's sure to fail before  
I've used it 6 full times? Because that's what your "1 per every 12  
TB" claim seems to imply. I don't think manufacturers would release  
drives with such poor reliability, because I don't think consumers  
would stand for it.

Harris doesn't cite his URE figures - at least not in the "RAID5 2009"  
article, I'm not going to waste my time or give the bloke any more  
credit by reading this follow up. Harris is probably yakking it up  
about all his successful trolling and how he makes a nice living on  
the basis of it.

I'm not using any Dell RAID controllers extensively right now, but my  
3ware comes with a weekly schedule for verifying the arrays. I run the  
verification every day on my RAID5. If you can't afford the  
performance penalty of verifying daily, then you should probably be  
using newer hardware that supports RAID6 or better.

Previous comments I've made on the Harris article:

I'm completely open to accepting such conclusions as Harris makes, but  
someone needs to prove them first.


More information about the Linux-PowerEdge mailing list