This repository contains scripts for creating and managing incremental backups using an encrypted BTRFS filesystem as target.

The backup.sh script is designed to backup source filesystems that don't use BTRFS. See README-mixed.md for details.

In contrast to that the backup.py script is for backing up BTRFS to BTRFS. The following sections describe it in detail.

Georg Sauthoff mail@georg.so

Create basic configuration in ~/.config/btrarch.json:

{
  "destination" : {
    "device" : "/dev/disk/by-id/usb-some-id",
    "mapper_name" : "backup",
    "mount_point" : "/mnt/backup"
  },
  "source" : [
    {
      "path"         : "/home",
      "name"         : "home",
      "snapshot_dir" : "/snapshot",
      "destination"  : "/mnt/backup/example.org"
    },
    {
      "path"         : "/",
      "name"         : "slash",
      "snapshot_dir" : "/snapshot",
      "destination"  : "/mnt/backup/example.org"
    }
  ]
}

Retention plan is the default one.

Format the destination device:

# backup.py --init

Create first full backup:

# backup.py

Create next incremental backup and possibly remove outdated snapshots according to the retention plan:

# backup.py

BTRFS is a copy on write file system that supports fast snapshotting. Thus, using it for incremental backups suggests itself. For encryption, the BTRFS filesystem is created on a luks-encrypted device-mapper device.

Advantages:

• speed
• easy retrieval - the restore of the last or any previous snapshots can be done via simple filesystem commands. No need to restore several increments on each other or to construct some kind of catalogue.
• data integrity - since BTRFS checksums all filesystem data, errors are detected. The checksums are verified during normal filesystem operation - but it is also possible to explicitly verify a complete volume (cf. btrfs-scrub(8)).

Old local and remote snapshots are removed according to a retention plan. The default one keeps 1 snapshot per day for the last 7 days, after that, for 4 weeks 1 per week, after that, for 6 months, 1 per month and after that for 2 years 1 per year.

A custom retention plan can be specified in the JSON configuration file.

In case the number of snapshots in an interval is greater than specified in the plan, superfluous ones are randomly selected and removed.

Snapshots are kept locally and remotely. The main reason for keeping them also locally is convenience. All local snapshots except the last one can be removed to free up space. The last one is sufficient for the next incremental backup run.

Doing an incremental backup via btrfs send/receive is very fast because the filesystem is designed to efficiently support those operations. In contrast, on a traditional filesystem, tools like rsync have to traverse the complete directory tree and stat each file.

For example, a typical incremental backup of a 250 GB SSD disk (containing 2 BTRFS volumes), of lightly changed data, to a USB 3 external 2.5" spinning disk takes ~ 40 seconds. Including entering a secure password (which takes 5 seconds or so).

After some runs the local snapshot directory hierarchy might look like this:

/snapshot
├── home
│   ├── 2016-11-28T22:37:49.576739
│   ├── 2016-12-02T22:52:22.110225
│   ├── 2017-01-30T23:42:22.647334
│   ├── 2017-02-06T12:23:34.267908
│   └── 2017-02-07T19:32:40.919178
└── slash
├── 2016-12-05T09:11:39.474245
├── 2016-12-11T23:47:14.647986
├── 2017-02-06T12:23:34.267908
└── 2017-02-07T19:32:40.919178

As expected, the structure is similar on the backup device, e.g.:

/mnt/backup/example.org/
├── home
│   ├── 2016-12-04T23:50:01.556331
│   ├── 2016-12-05T09:11:39.474245
│   ├── 2017-01-30T23:42:22.647334
│   ├── 2017-02-06T12:23:34.267908
│   └── 2017-02-07T19:32:40.919178
└── slash
├── 2016-12-05T09:11:39.474245
├── 2016-12-11T23:47:14.647986
├── 2017-02-06T12:23:34.267908
└── 2017-02-07T19:32:40.919178

In case BTRFS is not available as a filesystem (e.g. because the kernel is too old or the system isn't Linux) a good alternative is to use a solution based on Dar.

GPLv3+