Jack Tanner wrote:
I think I may have just painted myself into a corner. Could someone
who knows LVM and software RAID look this over?
I installed FC5 with the default partitioning settings (/boot + LVM
partition containing LVs for / and /home) on /dev/hda. Then I added
two more drives to the box, and did
# mdadm --create /dev/md0 -l raid1 -N 2 /dev/hdb /dev/hdc
Then I used system-config-lvm to join /dev/md0 to the existing volume
group, and moved the extent containing /home to reside on /dev/md0.
The idea was that I want my data on redundant disks, but I don't care
about the OS itself, because that can be trivially reinstalled.
Everything worked just fine.
Then I rebooted, and the kernel halted on boot ...
Redhat nash starting
Couldn't find device with uuid ...
Couldn't find all physical volumes for volume group VolGroup00
I can boot off a rescue CD, and do
# mdadm -A /dev/md0 /dev/hdb /dev/hdc
mdadm: /dev/md0 has been started with 2 drives
After that, I can see via lvm that all the VGs and LVs are there.
So it seems like /dev/md0 isn't getting assembled on boot up. I tried
following the mdadm man page and had grub pass the kernel the
parameter md=0,/dev/hdb,/dev/hdc , but that did absolutely nothing.
So, basically, I have two questions. First, how do I go about
diagnosing this and getting the box to boot up right?
Second, am I being an idiot in the first place by making my VG contain
a single drive with the OS and a software RAID1 array with /home? That
is, if the non-redundant OS drive fails, will I still be able to plop
the OS on a new drive, and join the RAID1 array (or even just one of
the two drives in the array) to the new drive's VG?
Two things come to mind that could cause that:
1) First, for Software RAID I've always created a partition and set it
to type fd rather than use the device itself.
2) If you create a RAID disk after installation, you need to create a
initrd image (mkinitrd) that supports software RAID.
Hope this helps,
Daniel
