STATUS //
Active
STACK //
Restic
Backrest
Backblaze B2
Encrypted off-site USB
Syncthing (sync only)
LAST UPDATED //
July 2026
Nexus eventually became important enough that "I'll just rebuild it"
stopped being the backup plan. After so many hours of tinkering and
configuring, it only made sense to put a proper backup plan in place.
The goal is not to back up every single replaceable media file. It is to
protect the configuration, service data, and notes that would be a pain
to reconstruct after a drive failure, bad update, or one of my own
mistakes.
The result is a layered setup, like a backup burger: Restic snapshots
managed through Backrest, encrypted cloud copies in Backblaze B2, and a
manual USB backup stored off-site.
BACKUP PHILOSOPHY //
A backup is only useful if I know what it contains, where it is stored,
and how to restore it.
I'd rather back up a smaller set of important data properly than blindly
copy everything and hope I can restore it later.
BACKUP LAYERS //
The first layer is an automated cloud backup. Backrest manages encrypted
Restic snapshots and sends them to Backblaze B2 every day at 4:30 AM
local time. The scope stays deliberately small: system config in /opt
and /etc, Docker volumes, and the notes and backups folders inside my
Nexus Sync directory. Retention keeps 7 daily snapshots, 4 weekly, and 3
monthly, with an automated integrity check and prune running once a
month on top of the daily backup itself.
The second layer is a manual encrypted USB backup. There is no fixed
schedule for it. I run it whenever I think of it, and it exists purely
as a failsafe: the drive, a Kingston DataTraveler, sits in secure
off-site storage the rest of the time, covering the same kind of
critical data as the cloud layer in case of a real disaster.
Syncthing helps move notes and other important files between devices,
but it is not treated as a backup. A deletion can sync just as easily as
a useful change, so those files still need to exist inside the actual
backup layers.
WHAT IS NOT BACKED UP //
Movies, TV shows, and other replaceable media are intentionally left out
of the backup plan. Rebuilding those libraries would be annoying, but
not nearly as annoying as losing weeks of configuration, notes, or
service data.
Backing up everything would increase storage use, transfer time, and
cost without improving recovery where it matters most. I do not need 15
copies of Akira stored across several drives. Keeping the scope this
tight also means the entire cloud backup comfortably fits inside
Backblaze B2's free 10 GB tier, which is a decent forcing function for
staying honest about what actually needs protecting.
I am also not the type of person to hoard a massive library of unwatched
media just to fill hard drive space. I keep a relatively small backlog,
keep tabs on it, and generally don't hang on to media after I've watched
it.
RESTORE TESTING //
A backup is only useful if I can actually get the data back. I
occasionally roleplay a disaster by restoring a small file or folder
into an isolated location, then verify the repository afterward.
Backrest's own monthly check and prune adds another layer on top of
that, validating the repository itself even on the months I don't
manually test a restore.
A little excessive? Possibly, but it gives me peace of mind instead of
assuming a successful backup job means recovery will always work.
The goal is not to torture myself by rehearsing a full disaster every
week. It is to know the backups are readable and that I understand the
restore process before I actually need it.