Re: kern.version

[Nick Holland <nick@holland-consulting.net>]

D. E. Evans wrote:
> Nick wrote:
>    Somehow, you got the GENERIC kernel on your second drive.  I
>    don't care how, but one example would be you installed your
>    system, you then mirrored your existing disk using the BIOS in
>    the (non)RAID controller, then booted OpenBSD, which promptly
>    ignored the "mirror".
> 
>    Your [insert rude, obnoxious, and not repeatable in polite
>    company words here] RAID controler has a BIOS that thinks it
>    can boot from whatever it darned well feels like, and does so
>    (in its defense, it probably thinks your mirrored pair is
>    corrupt (heh, it us right, actually, and the drive that was
>    "corrupted" WAS the wd0 drive, as it was written to "away"
>    from the mirror), and thinks the second drive is more likely
>    the "intact" one). OpenBSD then picks a drive as wd0, which
>    obviously isn't what the BIOS considered hd0, loading the
>    wrong kernel.
> 
>    Remember: until the machine boots, there is no wd0, that's an
>    OpenBSD concept.  The kernel is loaded by the BIOS, which can
>    do whatever it wants.
> 
> This seems a precise description.  It tooks Dan's verbal 2x4 to
> realize this, and your note to confirm it.
> 
> So is it worth learning the ins and outs of RAIDframe to make the
> RAID controller work, or is there a way I can get the promise
> controller to act like a regular EIDE drive?  (A software
> controlled RAID controller seems to defeat the whole point:
> RAID minus the controller.)

Hense my line of adjectives I figured I'd best leave to your imagination. :)

I've used a few of these controllers in non-RAID mode, by simply never 
establishing the RAID in the first place.  Now that you have the RAID 
"established" (at least, that's what the controller thinks), you can 
likely "nuke" that by 'dd'ing the whole drive with zeros, or more 
likely, you can probably just zero out the first and last couple sectors 
on the drive with dd (the systems I work with "mark" the drive in this 
way as part of a mirrored set.  The goal here is to remove the mark. 
'bout the only places to mark it are at the very beginning or very end 
of the drive).  Or your BIOS might have an option to remove both drives 
from the mirrored set, in which case, it will start to treat them as two 
individual drives, and you are in business.

If mirroring is an issue to you, I'd suggest the Accusys boxes.  They 
just work.

Nick.
-- 
http://www.holland-consulting.net