I need some more space in my computer, therefore I want some basic changes be done. One of those is setting a RAID up for all files to be a bit more robust (I have just had a hd crash, so I am a bit paranoid at the moment).
The old setup: One dedicated root and one boot partition on a HD. Nothing special. The other partitions are located under LVM (does not matter here).
The new (desired) setup:
The two partitions
/boot are merged into one single partition and managed under a RAID5. The LVM will be under another RAID5.
What I did: I created on all new HDs the partitions: One for the root RAID, one for swap and one for the rest. I created the root file system and copied the relevant data there. I modified
/etc/fstab to use the correct device (RAID partition and not mounting on
/boot). Then I cheated in the
grub.cfg of the old system such that I can boot into the new root. Just a quick hack to be able to boot up.
Now the problem arises: I rebooted and was able to start the new (RAID based) root file system. There I made a
update-grub and installed grub on another HD. So now I am able to select via the bios which grub version to use. If I select the old one, the RAID based root file system can be loaded and everything works. If I try this from the new grub version (automatically generated and using data only from the RAID partition), I end up in a grub rescue console. I get some error like:
error: No such device .... Here the UUID of the RAID Device is inserted.
So my question(s):
- Can grub generaly load its data from a raid device?
- Are there any special actions needed?
- Is it possible with and level and metadata version?
- Do I need a separate
/bootpartition to allow correct booting?
PS: It is a debian squeeze system.