-
-
Notifications
You must be signed in to change notification settings - Fork 61
Manual
If you're new to zfs-autobackup you should read the Getting started page first. It shows you how to get things done.
Then read this full manual to understand how zfs-autobackup works, and what all the options are doing.
Both guides are complementary.
usage: ZfsAutobackup.py [--help] [--test] [--verbose] [--debug] [--debug-output] [--progress] [--utc] [--version] [--ssh-config CONFIG-FILE] [--ssh-source USER@HOST]
[--ssh-target USER@HOST] [--property-format FORMAT] [--snapshot-format FORMAT] [--hold-format FORMAT] [--strip-path N] [--exclude-unchanged BYTES]
[--exclude-received] [--no-snapshot] [--pre-snapshot-cmd COMMAND] [--post-snapshot-cmd COMMAND] [--min-change BYTES] [--allow-empty] [--other-snapshots]
[--set-snapshot-properties PROPERTY=VALUE,...] [--no-guid-check] [--no-send] [--no-holds] [--clear-refreservation] [--clear-mountpoint]
[--filter-properties PROPERTY,...] [--set-properties PROPERTY=VALUE,...] [--rollback] [--force] [--destroy-incompatible] [--ignore-transfer-errors] [--decrypt]
[--encrypt] [--zfs-compressed] [--compress [TYPE]] [--rate DATARATE] [--buffer SIZE] [--send-pipe COMMAND] [--recv-pipe COMMAND] [--no-thinning]
[--keep-source SCHEDULE] [--keep-target SCHEDULE] [--destroy-missing SCHEDULE]
[BACKUP-NAME] [TARGET-PATH]
ZfsAutobackup.py v3.3-beta - (c)2022 E.H.Eefting (edwin@datux.nl)
positional arguments:
BACKUP-NAME Name of the backup to select
TARGET-PATH Target ZFS filesystem (optional)
Common options:
--help, -h show help
--test, --dry-run, -n
Dry run, dont change anything, just show what would be done (still does all read-only operations)
--verbose, -v verbose output
--debug, -d Show zfs commands that are executed, stops after an exception.
--debug-output Show zfs commands and their output/exit codes. (noisy)
--progress show zfs progress output. Enabled automaticly on ttys. (use --no-progress to disable)
--utc Use UTC instead of local time when dealing with timestamps for both formatting and parsing. To snapshot in an ISO 8601 compliant time format you may for example
specify --snapshot-format "{}-%Y-%m-%dT%H:%M:%SZ". Changing this parameter after-the-fact (existing snapshots) will cause their timestamps to be interpreted as a
different time than before.
--version Show version.
SSH options:
--ssh-config CONFIG-FILE
Custom ssh client config
--ssh-source USER@HOST
Source host to pull backup from.
--ssh-target USER@HOST
Target host to push backup to.
String formatting options:
--property-format FORMAT
Dataset selection string format. Default: autobackup:{}
--snapshot-format FORMAT
ZFS Snapshot string format. Default: {}-%Y%m%d%H%M%S
--hold-format FORMAT ZFS hold string format. Default: zfs_autobackup:{}
--strip-path N Number of directories to strip from target path.
Selection options:
--exclude-unchanged BYTES
Exclude datasets that have less than BYTES data changed since any last snapshot. (Use with proxmox HA replication)
--exclude-received Exclude datasets that have the origin of their autobackup: property as "received". This can avoid recursive replication between two backup partners.
Snapshot options:
--no-snapshot Don't create new snapshots (useful for finishing uncompleted backups, or cleanups)
--pre-snapshot-cmd COMMAND
Run COMMAND before snapshotting (can be used multiple times.
--post-snapshot-cmd COMMAND
Run COMMAND after snapshotting (can be used multiple times.
--min-change BYTES Only create snapshot if enough bytes are changed. (default 1)
--allow-empty If nothing has changed, still create empty snapshots. (Same as --min-change=0)
--other-snapshots Send over other snapshots as well, not just the ones created by this tool.
--set-snapshot-properties PROPERTY=VALUE,...
List of properties to set on the snapshot.
--no-guid-check Dont check guid of common snapshots. (faster)
Transfer options:
--no-send Don't transfer snapshots (useful for cleanups, or if you want a separate send-cronjob)
--no-holds Don't hold snapshots. (Faster. Allows you to destroy common snapshot.)
--clear-refreservation
Filter "refreservation" property. (recommended, saves space. same as --filter-properties refreservation)
--clear-mountpoint Set property canmount=noauto for new datasets. (recommended, prevents mount conflicts. same as --set-properties canmount=noauto)
--filter-properties PROPERTY,...
List of properties to "filter" when receiving filesystems. (you can still restore them with zfs inherit -S)
--set-properties PROPERTY=VALUE,...
List of propererties to override when receiving filesystems. (you can still restore them with zfs inherit -S)
--rollback Rollback changes to the latest target snapshot before starting. (normally you can prevent changes by setting the readonly property on the target_path to on)
--force, -F Use zfs -F option to force overwrite/rollback. (Useful with --strip-path=1, but use with care)
--destroy-incompatible
Destroy incompatible snapshots on target. Use with care! (also does rollback of dataset)
--ignore-transfer-errors
Ignore transfer errors (still checks if received filesystem exists. useful for acltype errors)
--decrypt Decrypt data before sending it over.
--encrypt Encrypt data after receiving it.
--zfs-compressed Transfer blocks that already have zfs-compression as-is.
Data transfer options:
--compress [TYPE] Use compression during transfer, defaults to zstd-fast if TYPE is not specified. (gzip, pigz-fast, pigz-slow, zstd-fast, zstd-slow, zstd-adapt, xz, lzo, lz4)
--rate DATARATE Limit data transfer rate in Bytes/sec (e.g. 128K. requires mbuffer.)
--buffer SIZE Add zfs send and recv buffers to smooth out IO bursts. (e.g. 128M. requires mbuffer)
--send-pipe COMMAND pipe zfs send output through COMMAND (can be used multiple times)
--recv-pipe COMMAND pipe zfs recv input through COMMAND (can be used multiple times)
Thinner options:
--no-thinning Do not destroy any snapshots.
--keep-source SCHEDULE
Thinning schedule for old source snapshots. Default: 10,1d1w,1w1m,1m1y
--keep-target SCHEDULE
Thinning schedule for old target snapshots. Default: 10,1d1w,1w1m,1m1y
--destroy-missing SCHEDULE
Destroy datasets on target that are missing on the source. Specify the time since the last snapshot, e.g: --destroy-missing 30d
zfs-autobackup uses safe defaults such as:
- Preserving all dataset properties. This can have its drawbacks to, for example Mounting
- Preserving full dataset paths. Look here how to change that.
- Only modify snapshots that match the zfs-autobackup format.
- Not rolling back or forcing anything.
- Checking everything, failing early and in a verbose manner.
- Don't do anything unexpected.
Keeping this in mind can help make more sense of the options described here: most of the options are to modify these safe defaults.
You probably already know these from the Getting started guide, scroll on to Step 1 in that case.
It's recommended to always use --verbose or -v to see whats going on. It makes debugging easier.
During initial setup and testing of a backup you should use --test. This will perform all the read-only operations, but will not change anything. It will show you exactly what it's going to do.
If you encounter a problem and want to see the exact ZFS commands, use --debug. This outputs all the underlying zfs-commands in a different color. To see the output of each command, use --debug-output.
Note that debug mode will abort on the first failed dataset, and show a stacktrace so it's not recommended for use in production.
zfs-autobackup can be used locally or remotely via ssh.
These options are for backing up to or from remote hosts via ssh:
-
--ssh-source USER@HOST: Source host to pull backup from. -
--ssh-target USER@HOST: Target host to push backup to.
If you dont specify a source or target host, zfs-autobackup will operate locally.
Things like different ssh-ports should be configured in your ~/.ssh/config file. (Or the one specified with --ssh-config CONFIG-FILE)
zfs-autobackup always performs its operations in a certain order.
All these steps are described here:
This step selects the datasets that are part of the run.
Selection is done by a dataset property. The name of this property is the backup-name, formatted by --property-format. The default is autobackup:backup-name.
The zfs-autobackup property can have the following values:
-
true: Select the dataset and all its children. -
false: Exclude the dataset and all its children. -
child: Only select the children of the dataset, not the dataset itself. -
parent: Only select the parent, but not the children. (supported in version 3.2 or higher)
If there are no datasets that have this property set then zfs-autobackup exits with an error.
Datasets can also be excluded from selection by these options:
-
--exclude-received: Exclude datasets that have the origin of theirautobackup:backup-nameproperty as "received". This can avoid recursive replication between two backup partners. -
--exclude-unchanged BYTES: Exclude datasets that have less than BYTES data changed since the last snapshot. (Use with proxmox HA replication)
In this step a snapshot is created on the datasets selected in step 1.
zfs-autobackup creates atomic snapshots per pool. This is a single zfs snapshot command that includes all the snapshots that need to be taken for that pool.
Snapshotting can be skipped with --no-snapshot. Using this option will result in only syncing existing snapshots.
Snapshots are created using a specific naming format. This includes a timestamp that zfs-autobackup uses to determine when a snapshot can be destroyed by the Thinner.
It is possible to change this format by using --snapshot-format. Other snapshots that do not match this format are normally ignored by zfs-autobackup. Use --utc to use UTC for timestamps.
You can run commands pre- and post-snapshotting with --pre-snapshot-cmd and --post-snapshot-cmd.
More info here.
Snapshot creation will be skipped for datasets that have no changes since the last snapshot.
This can be controlled by:
-
--min-change BYTES: Only create snapshot if enough bytes have changed. (default 1) -
--allow-empty: If nothing has changed, still create empty snapshots. (Same as --min-change=0)
-
--set-snapshot-properties PROPERTY=VALUE,...: List of properties to set on the new snapshot.
Syncronisation is done only if TARGET-PATH is specified. Otherwise zfs-autobackup is just a snapshot tool and stops after step 2.
For each selected source dataset it does the following steps:
If the target dataset already exists:
- We determine the Common snapshot
- We check the GUID of the common snaphot, unless
--no-guid-checkis set. - We determine a list of incompatible snapshots that are in the way. (after our common snapshot)
- If there isn't a valid common snapshot, this dataset fails and we continue with the next one.
We determine which snapshots are kept and which ones can be destroyed by the Thinner. Note that only our own snapshots (matching the --snapshot-format), are considered for deletion.
If --no-thinning is used, this list of obsolete snapshots will always be empty.
After planning, provided --no-thinning isn't used, we destroy obsolete snapshots on the source and target to save space during sync.
If the planner has detected incompatible snapshots, we will destroy them. But since this can be dangerous and is normally not needed, you have to enable this with --destroy-incompatible
Otherwise the dataset will fail.
Now the snapshots are actually transferred, unless --no-send is used.
If --other-snapshots is specified, we will also transfer snapshots that do not match our --snapshot-format. These other snapshots will never be destroyed.
For each snapshot we:
- Check if we need to resume an aborted transfer.
- Handle Encryption options. (
--encryptand--decrypt) - Transfer the data, using various Transfer options (
--zfs-compressed,--compress,--send-pipe,--recv-pipe,--buffer,--rate) - Filter/set properties according to
--set-propertiesand--filter-properties - Add/remove holds, unless
--no-holdsis used. (Use--hold-formatto specify the name of this hold)
Just before the first snapshot is transferred, we do a rollback, if --rollback is specified.
If it's an initial transfer that created a new target dataset, we try to automount the target after the first snapshot is transferred.
We destroy obsolete snapshots from the planning phase as soon as possible. (--no-thinning effectively disables this )
-
--ignore-transfer-errors: Ignore ZFS transfer errors. It still checks if received filesystem exists. This is useful to ignore some acltype errors. -
--clear-refreservation: Filter "refreservation" property. Recommended to save space. Same as--filter-propertiesrefreservation. -
--clear-mountpoint: Set propertycanmount=noautofor new datasets. Recommended, prevents mount conflicts. Same as--set-properties canmount=noauto. Also see Mounting -
--strip-path N: Number of directories to strip from target path. An example is given in the Getting started guide. -
--force: Usezfs -Foption to force overwrite/rollback. Useful with--strip-path=1. Use with care!
Datasets that are missing or deselected, but still exist in the target-path are called missing datasets.
The handling of those is described here (--destroy-missing )
The thinner decides when a snapshot is obsolete. Look at Thinner for more info. (--keep-source and --keep-target)
In order to run zfs-autobackup without root permissions, you'll need to set a few ZFS permissions. The permissions required differ for receiving and sending.
On the machine you want to sync the dataset from, you'll need these permissions:
root@source:~# zfs allow -u localuser mount,send,hold,snapshot,destroy,release rpoolOn the receiving side, you need these:
root@target:~# zfs allow -u remoteuser compression,mountpoint,create,mount,receive,rollback,destroy,release tank/backups/rpoolDepending on your use-case, you can always try to restrict these permissions further.
On some distributions like Alpine linux, you will need to change the permissions of /dev/zfs, so that regular users can write to it. This is safe, since the zfs kernel module handles access restrictions internally. (via zfs allow etc)
If you have trouble, make sure to check this, or just fix it with chmod 666 /dev/zfs
Sponsored by: JetBrains
zfs-autobackup:
- Introduction (README.md)
- Getting started
- Full manual
- Mounting backup datasets
- Performance tips (recommended)
- Common problems and errors
- Thinning out obsolete snapshots
- Common snapshot and holds
- Handling ZFS encryption
- Transfer buffering, compression and rate limiting.
- Custom Pre- and post-snapshot commands
- Monitoring
Examples: