Re: netbsd and RAID and UCR errors on large drives

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

Thilo Jeremias writes:
> Gents,
> 
> I'm wondering about the algorithm that is implemented in netbsd.
> I believe that the most common errors for large drives are single read 
> failures,
> those errors should be (IMHO) recoverable in a raid set by re-writing 
> the failing blocks
> from a working drive.
> 
> Discussions suggest, that the non decreasing error rate for drives, will 
> cause in a couple of years (assuming constant capacity increase),
> a guaranteed UCR while rebuilding a raid system, so the suggestion is to 
> use RAID6 to allow for that.

Perhaps in a couple of years all drives will be solid state, and we 
won't need to worry about them as much? ;)

> As I said above some drive failures should be avoidable w/o raid6 and 
> I'm wondering if
> netbsd does (or will) implement such a single block recovery ?

NetBSD currently does not implement such single block recovery.  
Doing so is on a TODO list, along with many other things (like 
working with the existing RAID 6 code).
 
> Is there another interface into the smart error list than smartctl ?

I don't know the answer to that one.

Later...

Greg Oster