Exercise 1: Rebuilding the Plan for a Bit-Rot-Protection Requirement — Possible Solution
====================================================================

Chapters that would change: fs1-3/fs1-4 (Btrfs/ZFS) and fs1-9
(the layering decision).

What would be dropped:

Per this chapter's own warn-box, "if bit-rot protection ever became an
actual requirement... Btrfs or ZFS (fs1-3/fs1-4) would be the correct
answer instead" of the ext4-on-LVM-on-mdadm stack actually built in
this capstone. Concretely, this means dropping the separate mdadm
layer (fs1-8) and the separate LVM layer (fs1-5) entirely -- per
fs1-9's own integrated-approach description, "Btrfs's own built-in
RAID profiles and ZFS's own RAID-Z... eliminate the separate RAID
layer entirely -- the filesystem is the volume manager is the
redundancy layer, all one system." There would be no /dev/md0 device
created with mdadm at all, and no separate pvcreate/vgcreate/lvcreate
step either, since Btrfs subvolumes or ZFS datasets (fs1-3/fs1-4) take
over the role LVM's logical volumes played in this capstone's own
stack.

What would replace it:

In place of Step 1 (mdadm --create) and Step 2 (pvcreate/vgcreate/
lvcreate/mkfs.ext4), a ZFS-based rebuild would use a single zpool
creation command establishing both the redundancy and the storage pool
at once, per fs1-4's own example: something like "zpool create tank
mirror /dev/sdb /dev/sdc mirror /dev/sdd /dev/sde" to build the
RAID-10-equivalent striped-mirror layout directly at the pool level,
followed by "zfs create tank/data" in place of the separate LVM/mkfs
steps.

What would stay the same:

Step 3 (a point-in-time snapshot before a risky step) and Step 4's own
underlying idea (simulate a disk failure, replace it, verify recovery)
both still apply conceptually, just using zfs snapshot and zpool
replace instead of lvcreate -s and the mdadm --fail/--remove/--add
sequence -- and this time, per fs1-3/fs1-4, the rebuild ("resilver" in
ZFS's own terminology) and every subsequent read would be checksummed
and self-healing, closing exactly the content-verification gap this
capstone's own warn-box named as a real, deliberate limitation of the
stack it actually built.

WHY THIS WORKS AS AN ANSWER
------------------------------
This identifies specifically which two layers (mdadm and LVM) get
eliminated and why, using fs1-9's own integrated-model explanation,
and names the concrete replacement commands from fs1-4 rather than
vaguely saying "use ZFS instead."