Testing RAID with loop devices

Of course. Thedev-loop? devices are part of the RAID (not the whole RAID) so each would only contain part of the filesystem...not enough for debugfs (or anything else) to make sense of it as a filesystem. You'd have better luck playing with the whole RAID device,dev-md0.

How do you know that? Where did you look? proc-mdstat will tell you all about the RAID's current status---which disks are active, which are failed, and whether or not it is re-syncing. In a properly configured RAID setup, the only way you should notice a failure is to see proc-mdstat showing a rebuild---which will be very fast on your small example RAID.

However, think about this: what happens when you delete an open file? Remember that the software RAID subsystem has your file open and won't close it until you shutdown the RAID. So the file won't really get deleted from disk while the RAID is up. You're better off using raidsetfaulty to tell the system that one of the RAID components is bad.

You're actually adding extra complexity with the loop devices and making it a non-real world example. It's both simpler and more realistic to get a bunch of cheap, old drives and use those, demonstrating what happens when the plug is pulled on one of them, IMHO. That more closely simulates the real world than manuall setting a loop device faulty with raidsetfaulty.

I also notice that you're usingetc-raidtab and mkraid---implying that you're using the old raidtools package. You may want to look into the newer mdadm instead.