Re: Thoughts on large disk reliability and raid maintenance?

[Greg Oster <oster%cs.usask.ca@localhost>]

Brian Buhrow writes:
>       Hello.  I've just been reading this blog on the reliability of larger
> and larger disks:
> 
> http://alumnit.ca/~apenwarr/log/?m=200809#08
> 
>       I wonder if it might be useful to add a thread which slowly and
> continuously reads the raid set over and over again, testing for bad
> sectors and, as he suggests, writing them back to the affected drives.
> I've certainly run into situations where a sector went bad on a raid set,
> then, when I went to rebuild the raid set, found that other sectors on
> other drives were bad.
> thoughts?

I was pointed to this article the other day... Let's just say that 
the claiming of a 56% error rate for a 1TB disk in the "short 
conclusion" didn't really align itself at all with the 56% error 
rate claimed for a 7x1TB RAID set in the article supposedly being 
summarized:

http://permabit.wordpress.com/2008/08/20/are-fibre-channel-and-scsi-drives-more-reliable/
 

At that point I pretty much backed up the salt truck...

"An Analysis of Data Corruption in the Storage Stack" by 
Bairavasundaram et al 
http://www.cs.wisc.edu/adsl/Publications/corruption-fast08.pdf

provides a more interesting (scary? :) ) read about disk data 
integrity if you're into that sort of thing...

(Yes, the idea of doing some sort of "srubbing" in RAIDframe has been 
thought of... no-one has gotten a round Tuit though... )

Later...

Greg Oster