> I have successfully set up a raid-1 array as my root partition
> using the latest tools and patches with kernel 2.2.7.
I assume 0.90 ...
> md: kicking non-fresh hdc3 from array!
> unbind<hdc3,1>
...
> md0: no spare disk to reconstruct array! -- continuing in degraded mode
I don't know the logic behind it, but the impression that I get is that a
failed (e.g. "non-fresh") is not deemed available for use as a spare disk ...
> Any ideas ??
raidhotremove /dev/md0 /dev/hdc3
raidhotadd /dev/md0 /dev/hdc3
> 2) The redhat rc.sysinit script trys to fsck the root partition twice,
> the second time when it's mounted. I have no idea why it does this
> for a /dev/md0 root and not a regular partition.
Boggle -- never tried it myself, but sounds wrong indeed ....
Sure /etc/fstab is right ?
> 3) I created seperate /boot partitions to contain System.map and vmlinuz.
> The are not mounted by default.
Why ?
> I also copied the files to /boot on the root partition
Are you **SURE** you got it right ?
One thing I remember from database lectures (several decades ago) is
One fact, One place
> but I still get tons of "/boot/System.map is not parsable as ..."
> messages when I start and shutdown.
[ I have a pet theory that there should be no need for a separate /boot in
the case od RAID1 ...
]
> 4) Just out of curiosity, does anyone know why to identical drives would
> come up with different geometries.
The BIOS ...
They tend to do more HACKs for the "first" disk than they do for others, as
one always boots off the first disk (!) so it is necessary to HACK round the
CHS disasters, etc.
> I think the BIOS is setting hda in tranlation mode, and leaving hdc alone.
Plausible.
Some HACK both masters, some just PM.
> Is there any way to force linux to overide the BIOS and use the real
> geometries on both?
Yes, several (why do you want to ?).
However, be careful if you want to use anything other than Linux ...
It is usually best to accept the defaults.
