Re: raidframe

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

On Wed, 25 Aug 2010 16:13:42 +0100
Patrick Welche <prlw1%cam.ac.uk@localhost> wrote:

> One of my raidframe mirror disks had unrecoverable errors. Running
> seatools fixed them (aside: would you return a 500G SATA disk if it
> had 64 errors?). The question now is, is the parity really alright?
> raidctl -p and -P say it is, but should I fail the disk and
> reconstruct it just to be sure? (There isn't a *really* verify parity
> option to raidctl is there?)

'raidctl -p' and 'raidctl -P' actually do the checking/rebuilding by
first comparing the blocks on each component, and then updating the
'mirror' if they are different.  If 'raidctl -p' or 'raidctl -P' have
just used the wonderful parity logging bits, then it's entirely
possible that your parity is still out-of sync on a few of the remapped
sectors (as those sectors will have been changed without the knowledge
of the parity mapping stuff). If 'raidctl -p' or 'raidctl -P' have
checked the entire disk, then you should be fine.  

If the disk with errors on it is the 'mirror' disk, then you can just
use 'raidctl -i' to re-initialize the parity.  It will go through the
entire disk and basically copy the 'primary' disk to the 'mirror' disk.

If the disk with errors on it is the 'primary' disk, then you're
probably best to do a 'raidctl -R' to reconstruct-in-place to the
primary (i.e. fail the primary, and then immediately rebuild to it).

Later...

Greg Oster