Documentation

Back Up to a NAS Without Losing Track of Versions

A NAS quickly accumulates three copies of the same folder with no one knowing which is authoritative. How to keep a usable record.

Preservation

The typical mess

A NAS starts clean, then fills up: Shoots_2024, Shoots_2024_v2, Shoots_2024_final, Shoots_2024_final_OK. Six months later nobody knows which is authoritative, and nobody dares delete.

That isn’t a tidiness problem. It’s a symptom: the NAS is serving as working volume, backup, and archive at once, without those three roles being distinguished.

Separate the three roles

  • The working volume holds what changes. It’s synced, fast, and is not a backup.
  • The backup holds dated states of the working volume, read-only, with an explicit retention policy.
  • The archive holds frozen, documented batches with a manifest and checksums. You never write to it again.

One rule is enough to hold the distinction: you don’t write into the archive. An archived batch is closed. If it needs correcting, you produce a new batch that supersedes it, without erasing the record.

Versioning, weakest to strongest

  1. Hand-suffixed folders. What everyone does, and what always fails: no reliable date, no policy, no space ever reclaimed.
  2. Filesystem snapshots. On a Btrfs or ZFS NAS, a snapshot is instant, cheap, and immutable. It’s the best effort-to-protection ratio available, and it’s often already present but not enabled.
  3. A backup with dated retention, run by a tool that applies a policy — keep the last seven days, the last four weeks, the last twelve months.
  4. Immutable archived batches, with a manifest, which no longer depend on any policy.

The RAID blind spot

Worth repeating: RAID is not a backup. It protects against a disk failing, not against deletion, not against propagated corruption, not against ransomware, not against fire. A RAID 6 NAS with no snapshots and no remote copy is a comfortable single copy.

In practice

  1. Physically separate work, backup, and archive — three distinct shares.
  2. Enable filesystem snapshots if available.
  3. Forbid writing to the archive technically, not just by convention.
  4. Name batches by identifier and date, never by final.
  5. Attach a manifest with checksums to every archived batch.
  6. Add an off-site copy: the NAS doesn’t survive the premises.

Obscura Flow builds documented batches with checksums to a NAS just as to object storage, and can re-check integrity from the destination — the archive stays identifiable and verifiable, without depending on a folder suffix.