Skip to content

pleiszenburg/abgleich

Repository files navigation

ABGLEICH

SYNOPSIS

abgleich is a simple ZFS sync tool. It displays source and target ZFS zpool, dataset and snapshot trees. It creates meaningful snapshots only if datasets have actually been changed. It compares a source zpool tree to a target, backup zpool tree. It pushes backups from a source to a target. It cleanes up older snapshots on the source side if they are present on the target side. It runs on a command line and produces nice, user-friendly, human-readable, colorized output. It also includes a GUI.

CLI EXAMPLE

demo

GUI EXAMPLE

snap backup cleanup
snap backup cleanup

INSTALLATION

The base CLI tool can be installed as follows:

pip install -vU abgleich

An installation also including a GUI can be triggered by running:

pip install -vU abgleich[gui]

Requires CPython 3.6 or later, a Unix shell and ssh. GUI support requires Qt5 in addition. Tested with OpenZFS 0.8.x on Linux.

abgleich, CPython and the Unix shell must only be installed on one of the involved systems. Any remote system will be contacted via ssh and provided with direct ZFS commands.

INITIALIZATION

All actions involving a remote host assume that ssh with public key authentication instead of passwords is correctly configured and working.

Let's assume that everything in source_tank/data and below should be synced with target_tank/some_backup/data. source_tank and target_tank are zpools. data is the "prefix" for the source zpool, some_backup/data is the corresponding "prefix" for the target zpool. For abgleich to work, source_tank/data and target_tank/some_backup must exist. target_tank/some_backup/data must not exist. The latter will be created by abgleich. It is highly recommended to set the mountpoint of target_tank/some_backup to none before running abgleich for the first time.

Rights to run the following commands are required:

command source target
zfs list x x
zfs get x x
zfs snapshot x
zfs send x
zfs receive x
zfs destroy x

config.yaml

Complete example configuration file:

source:
    zpool: tank_ssd
    prefix:
    host: localhost
    user:
    port:
target:
    zpool: tank_hdd
    prefix: BACKUP_SOMEMACHINE
    host: bigdata
    user: zfsadmin
    port:
include_root: yes
keep_snapshots: 2
keep_backlog: True
always_changed: no
written_threshold: 1048576
check_diff: yes
suffix: _backup
digits: 2
ignore:
    - home/user/CACHE
    - home/user/CCACHE
ssh:
    compression: no
    cipher: aes256-gcm@openssh.com
compatibility:
    target_samba_noshare: yes
    target_autosnapshot_ignore: yes

zpool defines the name of the zpools on source and target sides. The prefix value defines a "path" to a dataset underneath the zpool, so the name of the zpool itself is not part of the prefix. The prefix can be empty on either side. Prefixes can differ between source and target side. host specifies a value used by ssh. It does not have to be an actual host name. It can also be an alias from ssh's configuration. If a host is set to localhost, ssh wont be used and the user field can be left empty or omitted. Both source and target can be remote hosts or localhost at the same time. The port parameter specifies a custom ssh port. It can be left empty or omitted. ssh will then use its defaults or configuration to determine the correct port.

include_root indicates whether {zpool}{/{prefix}} should be included in all operations. keep_snapshots is an integer and must be greater or equal to 1. It specifies the number of snapshots that are kept per dataset on the source side when a cleanup operation is triggered. keep_backlog is either an integer or a boolean. It specifies if (or how many) snapshots are kept on the target side if the target side is cleaned. Snapshots that are part of the overlap with the source side are never considered for removal. suffix contains the name suffix for new snapshots.

Whether or not snapshots are generated is based on the following sequence of checks:

  • Dataset is ignored: NO
  • Dataset has no snapshot: YES
  • If the always_changed configuration option is set to yes: YES
  • If the tagging configuration option underneath compatibility is set to yes and the last snapshot of the dataset has not been tagged by abgleich as a backup: YES
  • written property of dataset equals 0: NO
  • Dataset is a volume: YES
  • If the written_threshold configuration is set and the written property of dataset is larger than the value of written_threshold: YES
  • If the check_diff configuration option is set to no: YES
  • If zfs diff produces any output relative to the last snapshot: YES
  • Otherwise: NO

Setting always_changed to yes causes abgleich to beliefe that all datasets have always changed since the last snapshot, completely ignoring what ZFS actually reports. No diff will be produced & checked for values of written lower than written_threshold. Checking diffs can be completely deactivated by setting check_diff to no.

digits specifies how many digits are used for a decimal number describing the n-th snapshot per dataset per day as part of the name of new snapshots. ignore lists stuff underneath the prefix which will be ignored by this tool, i.e. no snapshots, backups or cleanups.

ssh allows to fine-tune the speed of backups. In fast local networks, it is best to set compression to no because the compression is usually slowing down the transfer. However, for low-bandwidth transmissions, it makes sense to set it to yes. For significantly better speed in fast local networks, make sure that both the source and the target system support a common cipher, which is accelerated by AES-NI on both ends. The ssh port can be specified per side via the port configuration option, i.e. for source and/or target.

Custom pre- and post-processing can be applied after send and before receive per side via shell commands specified in the processing configuration option (underneath source and target). This can be useful for a custom transfer compression based on e.g. lzma or bzip2.

compatibility adds options for making abgleich more compatible with other tools. If target_samba_noshare is active, the sharesmb property will - as part of backup operations - be set to off for {zpool}{/{prefix}} on the target side, preventing sharing/exposing backup datasets by accident. If target_autosnapshot_ignore is active, the com.sun:auto-snapshot property will - similarly as part of backup operations - be set to false for {zpool}{/{prefix}} on the target side, telling zfs-auto-snapshot to ignore the dataset.

USAGE

All potentially changing or destructive actions are listed in detail before the user is asked to confirm them. None of the commands listed below create, change or destroy a zpool, dataset or snapshot on their own without the user's explicit consent.

abgleich tree config.yaml [source|target]

Show ZFS tree with snapshots, disk space and compression ratio. Append source or target (optional).

abgleich snap config.yaml

Determine which datasets on the source side have been changed since last snapshot. Generate snapshots on the source side where applicable.

abgleich compare config.yaml

Compare source ZFS tree with target ZFS tree. See what is missing where.

abgleich backup config.yaml

Send (new) datasets and new snapshots from source to target.

abgleich cleanup config.yaml [source|target]

Cleanup older local snapshots on source side if they are present on both sides. Of those snapshots present on both sides, keep at least keep_snapshots number of snapshots on source side. Or: Cleanup older snapshots on target side. Beyond the overlap with source, keep at least keep_backlog snapshots. If keep_backlog is False, all snapshots older than the overlap will be removed. If keep_backlog is True, no snapshots will be removed. If abgleich clean runs against the target side, an extra warning will be displayed and must be confirmed by the user before any dangerous actions are attempted.

abgleich wizard config.yaml

Runs a sequence of snap, backup and cleanup in a wizard GUI. This command is only available if abgleich was installed with GUI support.

SPEED

abgleich uses Python's type hints and enforces them with typeguard at runtime. It furthermore makes countless assertions.

The enforcement of types and assertions can be controlled through the PYTHONOPTIMIZE environment variable. If set to 0 (the implicit default value), all checks are activated. abgleich will run slow. For safety, this mode is highly recommended. For significantly higher speed, all type checks and most assertions can be deactivated by setting PYTHONOPTIMIZE to 1 or 2, e.g. PYTHONOPTIMIZE=1 abgleich tree config.yaml. This is not recommended. You may want to check if another tool or configuration has altered this environment variable by running echo $PYTHONOPTIMIZE.