David Hopper writes:
> Greg Oster wrote:
> > Is there an 'a' partition that automatically gets defined??!??!
> >
> > I'm curious to know that the (default) disklabels look like here...
> > (i.e. the output from 'disklabel sd1' and 'disklabel sd2')
> 
> Well, if I had actually _read_ the disklabel to begin with, I would have seen
> :
> 
> # /dev/rsd1c:
[snip]
> 3 partitions:
> #        size   offset     fstype   [fsize bsize   cpg]
>   c:  8380080        0     unused        0     0         # (Cyl.    0 - 3707)
> disklabel: boot block size 0
> disklabel: super block size 0

Ok, so it doesn't actually *have* a real label, just the faked-up one.

> I should add that this machine is coming from Digital UNIX 4.0E, and these 
> disks
> were never touched AFAIK, unless by DEC UNIX's automatic install.  Of course 
> there
> is no /dev/sd1a _or_ /dev/sd2a:  never was.  That's just me, thinking that 
> 'a'
> refers to the entire, unpartitioned disk, when every NetBSD doc under the sun
>  says
> it's 'c'.  (Okay, I'm done with the self-flagellation)  That said, raidctl _s
> hould_
> have given the above error,

Did 'raidctl -C ...' give any sort of error?  
(I realize the feedback from raidctl is (currently) rather weak... that
will hopefully change soon :) )

> but bear in mind that 'raidctl -C' means to force
>  the
> RAID creation _regardless_ of the disklabels.

'raidctl -C' only forces creation regardless of *component labels*, not
disklabels.  And 'raidctl -C' still requires there to be enough components
alive to actually start a RAID set.  If the /dev/sd1a doesn't exist, 
then when RAIDframe looks for that device, it shouldn't find it, and 
would mark it as 'failed'.  We're OK so far, since we can tollerate a
1-device failure in RAID 1.  When /dev/sd2a fails in the same way, however, 
sys/dev/raidframe/rf_disks.c:rf_ConfigureDisks() 
should fail, and print a message like:

   RAIDFRAME: Did not find any live disks in the array.

If it's not printing the above, then something weird must be happening 
during the lookups on /dev/sd1a and /dev/sd2a (maybe it's not even getting 
past the lookups??!.  (but all of this should still have nothing to do with 
bogus interrupts)

Later...

Greg Oster