More details can be provided, but here's the basis for the question. I just had a RAID-1 hardware failure which has resulted in an unusual situation:
Disk A, containing the file catalog data, but no file content data. The disk mounts and all files/folders are visible and valid and properly sized/attributed, but their contents are null/corrupt. I'm guessing the RAID broke down in the middle of a restore.
Disk B, containing the file content data, but no file catalog data. The disk mounts but appears as a blank/reformatted disk. With an undelete tool, the file contents are recoverable but without any names/folder structure/metadata (and many file types cannot be recovered).
As you may know, recovering raw data from a disk with no catalog is a horrendous undertaking, even with the proper tools.
So I'm wondering if it's possible to copy the catalog data from Disk A to Disk B, using a tool like dd
(low-level, exact-block-level-copy). Normally this would be out of the question, but I suspect it would be possible in theory if RAID-1 mirrored disks maintain identical data at the block level (same positions and lengths, not fragmented any differently). And keeping in mind that the catalog data for the same filesystem is always in the same place and format.
I just don't know if RAID-1 works this way by default, or if it throws blocks wherever it wants (like to avoid a bad block on one disk, or to optimize for seek times, putting the data in reverse order on each disk, etc).
I can of course make full disk backups with dd and try it out, but I wanted to know first since I'm working with a lot of data (~4TiB).
Another option: I've read that the NTFS filesystem stores a copy of the catalog in the middle of the volume. Unfortunately I'm using HFS+; does it do anything similar?
Update: Reading up a bit on Logical Block Addressing (LBA). It seems hard disks translate blocks to physical locations on the disk, which are arbitrary. So if two RAID-1 mirrors use identical block arrangements, then manually copying over known good blocks should be possible, even if the underlying magnetic bits are scattered all over each disk in different ways.
Update 2: Fixed the problem, though it was a bit involved. I'll post more details when I have a moment, in case it might help others who run into similar issues.