Borg is a client that creates
local backups, encrypts them, and saves them at a given location. A possible
location is the local filesystem, a mount of a remote storage (like CIFS),
certain storage protocols it directly can interact with or another borg binary.
The latter is the preferred option. You can place the borg binary on a remote
linux system and create an ssh account that cannot allocate a PTY and has
ForceCommand
set to borg serve
. Such u user can only send and receive
backups to the binary via ssh. A possible configuration for the sshd server
looks like this (won't be handled by this module):
# /etc/ssh/sshd_config
AcceptEnv LANG LC_*
ChallengeResponseAuthentication no
PasswordAuthentication no
PrintMotd no
Subsystem sftp /usr/libexec/sftp-server
UsePAM yes
X11Forwarding yes
Match Group borgusers
AllowAgentForwarding no
AllowTcpForwarding no
AuthorizedKeysFile %h/%u/.ssh/authorized_keys
ForceCommand borg serve
PasswordAuthentication no
PermitTTY no
PermitUserRC no
X11Forwarding no
This module will provide a borg backup script that works with such a setup. We will also configure a systemd service and a timer to execute it on a regular basis. Borg doesn't automatically prune old backups, our script has parameters for this.
Please have a look at our REFERENCE.d. All parameters are documented in that file.
The only parameter you really need to set is FQDN of the remote server:
class{'borg':
backupserver => 'myawesomebackupmachine.org'
}
We assume that your ssh username is the hostname from the client. You maybe want to overwrite this assumption:
class{'borg':
backupserver => 'myawesomebackupmachine.org',
username => 'notmyhostname',
}
We need to tell the script what we want to do with every mountpoint, backup it or ignore it. We can also exclude specific paths. The defaults are stored as hiera data in this module.
Figuring out from which backup archive you want to restore a certain file can be quite time-consuming with just borg alone. When listing the contents of each backup archive, the client will talk to the remote server a lot during the generation of the list. To speed this up, Florian Pritz developed a helper. This will be executed after every backup. The script talks to the server and fetches a list of all files from the last backup. The information are stored in a local sqlite database. You can do restores directly via this script. You can find the upstream documentation (including examples) at metacpan.org
More and more people use prometheus. We vendor a bash script that can provide you metrics about your backups in the prometheus format. They are written to disk and the node_exporter can collect them.
On CentOS 8, the PowerTools repository needs to be enabled by the user. Packages from EPEL8 require the repository but it's disabled by default. For more information see:
borgbackup before 1.1.17 didn't depend on packages from PowerTools so this worked by accident.
This module has several unit tests and linters configured. You can execute them by running:
bundle exec rake test
Detailed instructions are in the CONTRIBUTING.md file.
Contribution is fairly easy:
- Fork the module into your namespace
- Create a new branch
- Commit your bugfix or enhancement
- Write a test for it (maybe start with the test first)
- Create a pull request
Detailed instructions are in the CONTRIBUTING.md file.
This module was originally written by Tim Meusel. It's licensed with AGPL version 3.