Niloen Backup

Backups you can read, restore with stock tools, and prove restorable — to disk, cloud, and tape as equal targets.

Get started Why NBackup?

Niloen Backup (short: NBackup) is an open-source backup system written in Go. It orchestrates GNU tar and a compressor as child processes and produces immutable, self-describing artifacts that you can copy, inspect, and restore without NBackup installed — and it doesn’t just store your backups, it rehearses restoring them (nb drill) and pages you when one fails.

A backup administrator should be able to reason about backups by looking at a sequence of immutable daily backup runs rather than a database of chunks.

The design descends from Amanda — balanced multilevel scheduling, immutable daily artifacts, cycle-based safety — modernized: Amanda is tape-first, while NBackup treats local disk, virtual tape, and object stores (S3, GCS, Azure Blob) as equal targets, and makes the common modern shape — land fast on disk, then replicate offsite — a first-class operation. One static binary, driven by cron: no daemons, no database (the catalog is a cache one media scan rebuilds).

This is a first version — an early release under active development. The core is complete and exercised by its own recovery drills, but expect rough edges; see Status & limitations.

What makes it different

  • Recoverability is proven, not assumed. nb drill actually restores a risk-biased sample of your data and throws it away, delivering the “0 errors” digit of 3-2-1-1-0. An unattended failure is loud — it exits non-zero and can email or page you.

  • Backups you can read. Each run is exactly one immutable directory (or tape span, or set of objects) you can list and understand without any NBackup-specific tooling. A full restores with one pipe:

    zstd -dc 000000-app01-home-L0.tar.zst | tar -xf -

  • Recovery never requires NBackup. Archives are plain GNU tar streams piped through a stock compressor (and, optionally, gpg). The tools that wrote an archive are the tools that read it. NBackup orchestrates; it never invents a format only it can open.

  • Disk, cloud, and tape are equal. The same artifact lives unchanged on any medium. nb copy moves one run between media; nb sync mirrors a whole medium offsite. Land on fast local disk, then replicate to S3 or tape.

  • Capacity, not rotation. You give a medium a storage capacity and a cycle (how often each source is fully backed up). The planner chooses backup levels, full frequency, and retention to fit — no balancing knobs to tune.

  • Encryption that keeps copies interchangeable. A dump is encrypted once at the source (via gpg); verifying integrity and replicating offsite never need the key.

Is NBackup right for you?

NBackup deliberately trades storage efficiency for operational transparency: there is no cross-backup deduplication and no chunk store. If you back up many similar machines or keep long, dense snapshot histories, a chunk-store tool (restic, Borg, Kopia) will store the same data in far less space. What it can’t give you is what NBackup exists for: artifacts a human can read, restores that need no special tool or intact repository, first-class tape alongside disk and cloud, and drills that prove your backups restore. See How NBackup compares for an honest look at both sides.

How a backup is shaped

A volume is an ordered sequence of self-describing files. A run is a set of archives; each archive is its payload, a member index (its file list), and a commit footer (identity, sizes, checksums) written last — so the footer’s presence proves the archive landed whole. On disk it looks like this:

runs/run-2026-06-21.020000/
  000000-app01-home-L0.tar.zst        # clean compressed tar (payload)
  000000-app01-home-L0.hdr            # JSON header sidecar
  000001-app01-home-L0-index.json.gz  # gzipped member list (browse without extracting)
  000002-app01-home-L0-commit.json    # per-archive footer: identity + sizes + checksums
  ...

The same shape maps to an object store (one clean object per file) or tape (the header is a 32 KB block inline ahead of each payload). See Concepts for the full vocabulary and Artifacts you can read.

Where to go next

If you want to… Go to
Understand why NBackup is built this way Rationale
Install it and run your first backup Getting Started
Learn the vocabulary (run, DLE, cycle, …) Concepts
Read about a specific capability Features
Copy a working setup for your situation Scenarios
Look up a command or config key Reference
Come from Amanda Migrating from Amanda
Compare it to restic/Borg, Bacula, pgBackRest… NBackup vs the alternatives

NBackup is free software under the GNU General Public License v3.0, developed and maintained by Niloen AB, a software development consultancy. Copyright © 2026 Niloen AB. Bug reports and questions are welcome on GitHub issues; commercial support is available on request.