LVM vs RAID0 vs RAID "linear" - Combine 2 disks as one, data recovery?
If you must run multiple disks in a non-redundant configuration, and you want to be able to recover something when any of the disks fails, then, at the very least, you need to ensure that your data is not striped over all devices. When using a striped layout, parts of most files will end up on all disks, which means that when you lose any disk, you've irrevokably lost parts of nearly every file, not to mention parts of the filesystem structures themselves. This means, as you seem to already surmise, that you need to use linear allocation, which essentially appends the constituent devices' extents together, and gives you some chance of recovering files that exist entirely on the surviving device(s) after a device failure.
I could be wrong, but for LVM2, I believe linear allocation is the default when striping is not requested. Regardless, to be certain of a linear/appended layout, you can force it thus by first lvcreate
-ing a logical volume (LV) on the physical extents (PE's) of one disk, and then lvextend
-ing the LV onto the PE's of the other disk.
If you want to try software RAID (mdadm
), then I believe linear
mode is the one to use. Do not use raid0
, as it creates a striped layout, which, as I have argued, leads to almost total filesystem loss in the event of a device failure.
Finally, I should add that I don't recommend actually doing this and expecting to recover anything after a drive failure. If you're using LVM or mdadm to create a large scratch space out of smaller drives, then you'd better have your valuable data protected by other means (either backed up, or otherwise easily recreated).
You have an extremely high level of risk with combining two external 2TB USB drives into a 4TB configuration. First off, being external, the opportunity for physical disruption(unplugged, rough handling, etc) is considerable. Also many external drives employ power saving techniques that could end up trashing your Filesystem - especially in a LVM stripe / RAID 0 setup. In the case of disk failure, you will have to restore data from a backup.
A LVM stripe using both disk will give you zero recoverability.
Filesystem corruption - it depends on the extent on if the xxx.fsck tool can repair.
My suggestion - use RAID1.
You'll need to build a filesystem on each disk and mount them in different locations. Losing half a filesystem makes a consistent recovery of anything difficult.