The Neighbours
SnapRAID is one answer to "a disk died" among several, and it is not always the right one. This chapter lines up the alternatives and names what each trades. No neighbour was run in this guide's lab; these are design comparisons, with each project's own documentation as the place to verify specifics.
| Approach | Protects against | Costs | Watch out for |
|---|---|---|---|
| SnapRAID | Disk death, silent corruption, recent deletions | One disk per parity level; sync discipline | Nothing between syncs |
| Classic RAID 5/6 (mdadm, hardware) | Disk death, in real time | Matched-ish disks, all spinning | Rebuild failures; whole array at risk; not corruption-proof |
| ZFS (RAIDZ, mirrors) | Disk death, corruption, with snapshots for history | RAM and matched vdev planning | Growing an array is rigid; snapshots are not backups |
| Unraid | Disk death, one or two parity disks | A licence; its own OS | Real-time parity writes are slow; it's a platform choice, not a tool |
| Btrfs | Corruption detection, snapshots, RAID profiles | Less parity tooling than SnapRAID | Its RAID 5/6 has a long-documented write-hole history |
| Plain backup (rsync, restic, a second copy) | Everything, including fire and deletion-with-history | A full second copy's space | No help the day the only disk dies mid-week, if the backup is weekly |
The distinctions that matter
Real-time vs snapshot. RAID, ZFS and Unraid protect every write as it happens; SnapRAID protects at sync time. If your files change constantly and every version matters, real-time wins — and so does a real database backup, because none of these is one.
Corruption. Classic RAID happily mirrors corrupted data to every disk; it has no hashes. SnapRAID, ZFS and Btrfs hash blocks and can say which copy is right. On ageing disks this difference is the whole game.
History. Only snapshots and backups have a past. Parity of any flavour protects the present arrangement of bytes. Delete-and-sync, ransomware, a house fire — parity shrugs; a backup (or a snapshot you kept) answers.
Exit cost. SnapRAID and plain backup leave your disks as ordinary disks throughout. RAID and ZFS own the layout: leaving means copying everything off.
The usual honest answer
For the home case this guide targets, the common settled answer is SnapRAID for the shelf, plus a backup for the irreplaceable — parity absorbs the disk failure, the backup absorbs the mistakes and disasters. If you find yourself wanting real-time protection, checksums, snapshots and painless growth in one system, that list is ZFS's home ground and this guide will not pretend otherwise.
The one-sentence version
Parity answers "a disk died", snapshots answer "last Tuesday", backups answer "everything is gone" — pick the tools by which question keeps you up at night.