This script will check all datasets on a specified pool or pools, and where it
finds a local property specifying a destination host and pool, will send a
snapshot of that dataset to the destination. Called repeatedly, it will send
incremental snapshots optionally including any other snapshots created in the
interim.
It is safe to call as often as desired, even if a previous instance is already
running - in that case it will keep trying for a while, then give up and do
nothing if it can't get a lock.
To configure:
> zfs-mirror -r -s mypool zfs-backup:backup
Would set every dataset on pool 'mypool' to use pool 'backup' on machine
'zfs-backup' as its destination.
> zfs-mirror -s mypool/dataset zfs-backup:backup,zfs-backup-alternate:backup
Would set 'mypool/dataset' to use pools named 'backup' on both 'zfs-backup'
and 'zfs-backup-alternate' as its destinations.
To run:
> zfs-mirror
Would look for any datasets with any destinations configured, and
synchronise to them. Any datasets currently being mirrored to the same
destination by another version of this script will be skipped. This means it
can be run from cron as often as you like, without stacking up endless
copies of the script running in parallel.
> zfs-mirror -I -l -b 2000 mypool zfs-backup-alternate:backup
Would look for any datasets on 'mypool' which have
'zfs-backup-alternate:backup' configured as a destination, and synchronise
them at a maximum of 2000kBps, transferring any intermediate snapshots
between the ones created by this script (eg manually created or from
zfs-auto-snapshot), and saving a log to
'/var/log/zfs-mirror-est-send-sizes-mypool-to-zfs-backup-alternate_backup.log'
================================================================================
Usage: zfs-mirror [options] [source pool(s)] [destination pool(s)]
Source should be specified in the form "pool[,pool...]".
Default is all pools found on the local system.
Destination should be specified in the form "hostname:pool[,hostname:pool...]"
Default is all destinations configured for any datasets on the specified
source pool.
Available options:
-a <age>
Age report warning threshold: '-R' will warn for snapshots older than <age>.
Must be a number, optionally suffixed with d, h, m, or s.
Default is 86400 (24h)
-b <number>
Limit send rate to <number>kBps.
-c <option>
Colour: must be one of 'auto', 'never', 'always'.
Default is 'auto': colourise logging if output destination is a terminal.
-D Dedup: use deduplicated zfs send streams, as per 'zfs send -D ...'
-f Force: if an incremental mirror fails because the destination dataset has
been modified, determine the most recent common snapshot and use 'zfs
receive -F' to force the receiving side to roll back to it.
-g [source dataset(s)]
Get the destinations configured for the specified dataset(s), or all
configured datasets if none given.
May be combined with '-r' to include child datasets.
-h Help: display this usage note.
-I Send incremental streams using 'zfs send -I' rather than 'zfs send -i'
(See 'man zfs' for details.)
-k <number>
Keep <number> old snapshots on each destination dataset.
-l Log each transfer along with its estimated size in TSV format to
"/var/log/zfs-mirror-est-send-sizes-<source>-to-<destination>.log"
-n Dry run: don't make any changes, but print the commands that would have
been executed.
-o <option>
Output destination: must be one of 'auto', 'syslog', 'tty'
Default is 'auto': log to stderr if it is a terminal, else to syslog.
-p Display a progress bar for transfers. The total transfer size is an
estimate, so the progress display can be somewhat inaccurate; hence,
transfers estimated at < 5MB will not show a progress bar.
-R Report on the age of existing backups: for each dataset to mirror, produce
a warning if the most recent snapshot is older than a configurable
threshold (see '-a').
-r Recursive: Send child datasets if the destination property is inherited,
rather than requiring explicit configuration per-dataset.
-s Set destination: Configure each given source dataset to use the given
destination pool(s) as their mirror target(s). Both source and
destination must be specified.
May be combined with '-r' to set the destination for child datasets.
-t Use a temporary file on the destination rather than piping directly from
'zfs send' to 'zfs receive'. Requires 'zfs-ssh-wrapper' on the
destination.
-v Verbose: May be specified up to four times to increase the output
verbosity. By default only warnings and errors are printed.
-w Wait for lock: If unable to lock the source/dest pools, wait indefinitely
rather than giving up after 60s. This uses polling, so it is unlikely
that multiple waiting processes will proceed in FIFO order.
-z Use xz if available to compress transfers.