Exercise 2: Sizing a Snapshot for a 6-Hour, 2GB/Hour Backup Window — Possible Solution
====================================================================

Recommended size: at least 12GB reserved, with real headroom on top --
e.g. 18-20GB, not exactly 12GB.

Explanation:

Per the chapter's own warn-box, "the -L size on an LVM snapshot needs
to cover the total amount of data expected to change on the origin
during the snapshot's entire lifetime -- not the origin's total size."
The origin's total size (200GB) is explicitly the WRONG number to size
against here -- what matters is how much data will actually change
while the snapshot exists, which is a function of the write rate and
how long the snapshot needs to stay alive, not how big the database
is overall.

At 2GB of writes per hour over a planned 6-hour backup window, the
baseline expected change volume is 2GB x 6 = 12GB. Sizing the snapshot
at exactly 12GB, however, leaves no margin at all -- per the chapter's
own warn-box, "a heavily-written database left snapshotted for hours
can overflow a snapshot that seemed generously sized, silently
invalidating it." A 6-hour backup window is an estimate, not a
guarantee (the backup could run long, or the database's write rate
could spike above its usual 2GB/hour average during that window for
reasons unrelated to the backup itself), and per the chapter, an
overflow doesn't degrade gracefully -- it "silently invalidates" the
entire snapshot, losing the point-in-time state the backup depended on.

A reserved size with real headroom -- roughly 1.5x to 2x the baseline
estimate, e.g. 18-20GB -- absorbs a slower-than-expected backup run or
a temporary spike in write activity without risking an overflow that
would otherwise force the entire backup to be redone from a fresh
snapshot.

WHY THIS WORKS AS AN ANSWER
------------------------------
This sizes against the chapter's own stated correct basis (expected
change volume over the snapshot's lifetime, not total data size),
computes the baseline from the given numbers, and then explains why
that exact baseline is still too risky per the chapter's own overflow
warning, recommending headroom rather than a bare minimum.