Configuration¶
Benji only needs to be configured once.
benji.yaml¶
There is an example configuration which lists all possible configuration options:
configurationVersion: '1'
# Logfile location (default: null)
# Setting this to null disables logging to a file, this might be useful when running Benji inside a container.
# Benji will by default log INFO, WARNING and ERROR to this file. If you also need DEBUG information, please start
# Benji with the '-v' option.
logFile: /var/log/benji.log
# Default block size (default: 4194304)
# The block size can be changed on the command line on a version by version basis, but be aware that this will affect
# deduplication and increase the space usage. One possible use case for different block sizes would be backing up LVM
# volumes and Ceph images with the same Benji installation. While for Ceph 4MB is usually the best size, LVM volume
# might profit from a smaller block size.
blockSize: 4194304
# Hash function to use for calculating block checksums (default: BLAKE2b,digest_bits=256)
# There is normally no reason to change the default.
# DO NOT CHANGE WHEN BACKUPS EXIST!
hashFunction: BLAKE2b,digest_bits=256
# Process name (default: benji)
# This name will be used to identify a specific configuration in the
# process list and could be used to distinguish several parallel installations.
processName: benji
# Days of protection for backup versions (default: 6)
# Disallow removal of backup versions if they are younger than this number of days. Set to 0 to disable, i.e. to be
# able to delete any version.
disallowRemoveWhenYounger: 6
# Database engine (default: none)
# See https://docs.sqlalchemy.org/en/latest/core/engines.html for options.
# Only postgresql (dialect psycopg2) and sqlite are tested with during development.
# Examples: postgresql://scott:tiger@localhost/mydatabase',
# sqlite:////var/lib/benji/benji.sqlite
metadataEngine: sqlite:////var/lib/benji/benji.sqlite
# Default storage to use (no default, required)
defaultStorage: s-1
storage:
# Which data backend to use?
# Available types:
# file, s3, b2
type: file
file:
# Store data to this path. A structure of 2 folders depth will be created
# in this path (e.g. '0a/33'). Blocks of DEFAULTS.block_size will be stored
# there. This is your backup storage!
path: /var/lib/benji/data
#s3:
# awsAccessKeyId: ********
# awsSecretAccessKey: ********
# regionName: (optional)
# endpointUrl: (optional)
# signatureVersion: (optional)
# useSsl: true or false (default is true, ignored when endpointUrl is set)
# bucketName: benji
# multiDelete: true or false (default is true)
# addressingStyle: path, virtual or auto (default is path)
# disableEncodingType: true or false (default is false)
# activeEncryption: k1
# activeCompression: zstd
#b2:
# accountId: ***********
# applicationKey: *************
# bucketName: benji
# accountInfoFile: .... (optional)
# writeObjectAttempts: 1
# uploadAttempts: 5
# Encryption support
#
# The identifier is saved together with the object and the corresponding materials
# are then looked up during decryption. This means that you mustn't change the
# materials identified by a specific identifier after objects have been encrypted
# with it. If you want to change the encryption configurarion you must define
# a new encryption method with a new identifier. If you mark this encryption method
# as active it will be used to encrypt new objects. Reencrypting old objects is
# currently not implemented but it wouldn't be a problem to implement proper key
# rotation.
#
# Regarding kdfSalt and kdfIterations:
# It is highly recommended to generate your own random salt for example with:
# dd if=/dev/urandom bs=16 count=1 status=none | base64
# Don't change the salt and iteration count after writing encrypted objects
# to the data backend, they cannot be decrypted anymore.
#
# Specify either a password, kdfSalt and kdfIteraions or use your
# own high-entropy-key directly with masterKey in which case the other options
# are ignored.
#
#
# encryption:
# - identifier: k1
# type: aes_256_gcm
# materials:
# masterKey: !!binary |
# e/i1X4NsuT9k+FIVe2kd3vtHVkzZsbeYv35XQJeV8nA=
# OR
# password: "your secret password"
# kdfSalt: !!binary /YkKwy9InZVqC4p+eAdhlA==
# kdfIterations: 16384
# Currently only zstd compression is supported. You can specify the compression level:
# Higher values mean better compression ratios but increased CPU use. For details please
# see the zstd documentation.
#
# compression:
# - type: zstd
# materials:
# level: 3
# How many writes to perform in parallel. This is useful if your data backend
# can perform parallel writes faster than serial ones.
simultaneousWrites: 5
# How many reads to perform in parallel. This is useful if your data backend
# can perform parallel reads faster than serial ones.
simultaneousReads: 5
# Bandwidth throttling (set to 0 to disable, i.e. use full bandwidth)
# bytes per second
#bandwidthRead: 78643200
#bandwidthWrite: 78643200
#readCache:
# directory: /var/lib/benji/read-cache
# maximumSize: (in bytes)
nbd:
cacheDirectory: /tmp
io:
file:
# Configure the file IO (file://<path>)
# This is for a file or a blockdevice (e.g. /dev/sda)
# How many parallel reads are permitted? (also affects the queue length)
simultaneousReads: 5
rbd:
# Configure the rbd IO (rbd://<pool>/<imagename>[@<snapshotname>])
# This accepts rbd images in the form rbd://pool/image@snapshot or rbd://pool/image
cephConfigFile: /etc/ceph/ceph.conf
# How many parallel reads are permitted? (also affects the queue length)
simultaneousReads: 10
# When restoring images, new images are created (if you don't --force). For these
# newly created images, use these features:
newImageFeatures:
- RBD_FEATURE_LAYERING
- RBD_FEATURE_EXCLUSIVE_LOCK
#- RBD_FEATURE_STRIPINGV2
#-RBD_FEATURE_OBJECT_MAP
#-RBD_FEATURE_FAST_DIFF
#-RBD_FEATURE_DEEP_FLATTEN
clientIdentifier: admin
Configuration File Location¶
Benji will by default search the following locations for configuration files:
- /etc/benji.yaml
- /etc/benji/benji.yaml
- ~/.benji.yaml
- ~/benji.yaml
If multiple of these files exist, only the first file found is read.
In order to explicitly pass a configuration file, use the -c
(or
--configfile
) parameter.
Multiple Instance Installations¶
You can run Benji multiple times on different machines or in different containers simultaneously. The configurations will have to match! this is the responsibility of the user and isn’t checked by Benji. Be careful to shutdown all instances before making configuration changes that could affect other instances (like adding an encryption key).
Multiple instances open up the possibility to scale-out Benji for performance reasons, to put instances where the backup source data is or to have a dedicated instance for restores for example.
Locking between different instances is done via the central database.