On 11/3/26 19:25, Greg A. Woods wrote:
Using LVM and the device mapper to implement something equivalent to the
linux "dm-integrity" feature would be a great idea! It would be far
more flexible than just using RAIDframe.
I think doing integrity checks at the block level has the problem that
it can become far less actionable:
If BTRFS or ZFS complain that a file is broken, via an IO error trying
to read from it, I can just delete the file and copy it from a backup.
If a block layer complains a block is broken, I do... what? Toss the
entire install?
(I understand in the long term even a single bit flip is a strong
indicator for replacing the entire disk. But I prefer having both some
time and some room for nuanced and well-thought-out decisions when a
problem like this occurs, rather than having my hand forced.)
Regards,
Ellie
