Aleksandar Filipovski opens with a small disaster: the family photos, consolidated onto one external drive to free up disk space, then reformatted by his dad when a set-top box asked for it nicely. He refuses to blame him. One copy, no backup, no warning UI, and a non-technical user asked to make a storage decision — that is the design’s fault, not his.

The rest of the essay is the ladder that “just back up my files” turns into once you take it seriously:

  • A backup means a copy elsewhere — but a mirror is not one, because ransomware, a wrong delete or a bad script all propagate. RAID 1 is out; you need snapshots
  • Snapshots force a Recovery Point Objective: how much data can you afford to lose? Under 30 seconds for critical finance, a day or a week for everyone else
  • Rotation follows, because 365 snapshots a year is storage you can’t justify — daily/weekly/monthly (GFS), with granularity falling off as you go back in time
  • File changes are fat-tailed, so most files never change between snapshots — which argues for deduplication and hard links (what rsnapshot does), saving both disk and the bandwidth you pay cloud rates for
  • Your homelab then breaks it anyway: Docker containers write root-owned files that a default-user cronjob can’t read, and databases flush to disk in batches, so a filesystem-level copy restores corrupt data
  • Drive models with famous failure rates and physical risk push you to 3-2-1 — three copies, two media types, one offsite
  • Object storage punishes you twice: metadata is stripped on upload and many small objects are expensive, which drags you into tarballs and safe 50MB chunking. This, he says, is where he’d give up

The conclusion is a recommendation rather than a build log: use Borg or Restic, which already do encryption, chunk-level dedup and checksums, and thank the people who live-tested that trial and error so you don’t have to. The step everyone skips is the one that makes the rest real — actually restore from your backups, roughly every six months.

The appendix the discussion thread produced is nearly as good as the body: never schedule anything at or between 2 and 3 AM; note ZFS, a filesystem built around integrity and snapshots; and note consistency groups, because a snapshot taken while an application and its database are mid-write is not a backup you can restore.

The real target of the piece is the phrase “this is an afternoon job.”