pyznap is a ZFS snapshot management tool. It automatically takes and deletes snapshots and can send them to different backup locations. You can specify a policy for a given filesystem in the pyznap.conf file and then use cron to let it run regularly. pyznap includes zfs bindings for python, forked and modified from https://bitbucket.org/stevedrake/weir/.
pyznap regularly takes and deletes snapshots according to a specified policy. You can take frequent, hourly, daily, weekly, monthly and yearly snapshots. 'frequent' snapshots can be taken up to once per minute, the frequency can be adjusted by the cronjob frequency. Old snapshots are deleted as you take new ones, thinning out the history as it gets older.
Datasets can also be replicated to other pools on the same system or remotely over ssh. After an initial sync, backups will be done incrementally as long as there are common snapshots between the source and the destination.
pyznap is written in python 3.5+ and has no dependencies. For developing and running the tests you need:
pytest
pytest-dependency
paramiko
You also need the faketime
program for some tests to simulate pyznap running over time.
I suggest installing virtualenv & virtualenvwrapper, so you don't clutter your system python installation with additional packages.
pyznap uses mbuffer
and lzop
(by default) to speed up zfs send/recv, and pv
to show progress,
but also works if they are not installed. Other supported compression methods are: none
, lz4
,
gzip
, pigz
, bzip2
and xz
.
Note that ZFS needs root access to run commands. Due to this you should install pyznap under your root user.
pyznap can easily be installed with pip. In your virtualenv just run
pip install pyznap
and pyznap & its requirements will be installed. This should also create an executable in your PATH,
either at /path/to/virtualenv/pyznap/bin/pyznap
or /usr/local/bin/pyznap
. If you use your
system python installation you might want to use the --user
flag. In this case the executable will
be located at ~/.local/bin/pyznap
.
Before you can use pyznap, you will need to create a config file. For initial setup run
pyznap setup [-p PATH]
This will create a directory PATH
(default is /etc/pyznap/
) and copy a sample config there. A
config for your system might look like this (remove the comments):
[rpool/filesystem]
frequent = 4 # Keep 4 frequent snapshots
hourly = 24 # Keep 24 hourly snapshots
daily = 7 # Keep 7 daily snapshots
weekly = 4 # Keep 4 weekly snapshots
monthly = 6 # Keep 6 monthly snapshots
yearly = 1 # Keep 1 yearly snapshot
snap = yes # Take snapshots on this filesystem
clean = yes # Delete old snapshots on this filesystem
dest = backup/filesystem # Backup this filesystem on this location
exclude = rpool/filesystem/data/* # Exclude these datasets for pyznap send
Then set up a cronjob by creating a file under /etc/cron.d/
nano /etc/cron.d/pyznap
and let pyznap run regularly by adding the following lines
SHELL=/bin/sh
PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin
*/15 * * * * root /path/to/pyznap snap >> /var/log/pyznap.log 2>&1
This will run pyznap every quarter hour to take and delete snapshots. 'frequent' snapshots can be taken up to once per minute, so adjust your cronjob accordingly.
If you also want to send your filesystems to another location you can add a line
0 0 * * * root /path/to/pyznap send >> /var/log/pyznap.log 2>&1
This will backup your data once per day at 12am.
You can also manage, send to and pull from remote ssh locations. Always specify ssh locations with
ssh:port:user@host:rpool/data
If you omit the port (ssh::user@host:...
) the default 22
is assumed. A sample config which backs
up a filesystem to a remote location looks like:
[rpool/data]
hourly = 24
snap = yes
clean = yes
dest = ssh:22:user@host:backup/data # Specify ssh destination
dest_keys = /home/user/.ssh/id_rsa # Provide key for ssh login. If none given, look in home dir
compress = gzip # Use gzip compression for sending over ssh
To pull a filesystem from a remote location use:
[ssh::user@host:rpool/data] # Specify ssh source
key = /home/user/.ssh/id_rsa # Provide key for ssh login. If none given, look in home dir
dest = tank/data
compress = lz4
You can exclude datasets when sending using Unix shell-style wildcards:
[rpool]
dest = backup/rpool, tank/rpool
exclude = rpool/data rpool/home/*/documents rpool/home/user1, rpool/home*
Note that exclude rules are separated by a ,
for the different dests, and you can specify multiple
rules for a single dest by separating them with a space. Exclude rules thus cannot contain any comma
or whitespace.
I would also suggest making sure that root has ownership for all files, s.t. no user can modify them. If that is not the case just run
chown root:root -R /etc/pyznap/
Here is a list of all options you can set in the config fie:
Option | Input | Description |
---|---|---|
key |
String | Path to ssh keyfile for source |
frequent |
Integer | Number of frequent snapshots. These can be created every minute (whenever pyznap is called) |
hourly |
Integer | Number of hourly snapshots |
daily |
Integer | Number of daily snapshots |
weekly |
Integer | Number of weekly snapshots |
monthly |
Integer | Number of monthly snapshots |
yearly |
Integer | Number of yearly snapshots |
snap |
yes/no | Should snapshots be taken |
clean |
yes/no | Should snapshots be cleaned |
dest |
List of string | Comma-separated list of destinations where to send source filesystem |
dest_key |
List of string | Path to ssh keyfile for dest. Comma-separated list for multiple dest |
compress |
List of string | Compression to use over ssh, supported are gzip, lzop, bzip2, pigz, xz & lz4. Default is lzop. Comma-separated list for multiple dest |
exclude |
List of string | Whitespace-separated list of datasets to exclude from sending. Exclude lists for different dests are separated by comma |
raw_send |
List of yes/no | Use zfs raw send. Comma-separated list for multiple dest |
resume |
List of yes/no | Use resumable send/receive. Comma-separated list for multiple dest |
dest_auto_create |
List of yes/no | Automatically create missing root datasets. Comma-separated list for multiple dest |
retries |
List of integer | Number of retries on connection issues. Comma-separated list for multiple dest |
retry_interval |
List of integer | Time in seconds between retries. Comma-separated list for multiple dest |
Run pyznap -h
to see all available options.
-
--config
Specify config file. Default is
/etc/pyznap/pyznap.conf
. -
-v, --versbose
Print more verbose output.
-
setup [-p PATH]
Initial setup. Creates a config dir and puts a sample config file there. You can specify the path to the config dir with the
-p
flag, default is/etc/pyznap/
. -
snap
Interface to the snapshot management tool. Has three optional arguments:
-
--take
Takes snapshots according to policy in the config file.
-
--clean
Deletes old snapshots according to policy.
-
--full
First takes snapshots, then deletes old ones. Default when no other option is given.
-
-
send
Interface to the zfs send/receive tool. Has two usages:
-
No further option is given
Send snapshots to backup locations according to policy.
-
-s SOURCE -d DESTINATION [-c COMPRESSION] [-i KEYFILE] [-j SOURCE_KEY] [-k DEST_KEY] [-e EXCLUDE] [-w] [-r] [--dest-auto-create] [--retries RETRIES] [--retry-interval RETRY_INTERVAL]
Send source filesystem to destination filesystem. If either source OR dest is a remote location, you can specify the keyfile with the
-i
flag. If both source AND dest are remote, you specify the keyfiles with the flag-j
for the source and-k
for the dest. You can also turn on compression with the-c
flag. Currently supported options are:none
,lzop
,lz4
,gzip
,pigz
,bzip2
andxz
. If no option is given,lzop
is used if available. You can specify multiple (whitespace separated) wildcard exclude rules with the-e
flag. Note that you should probably pass these as strings or escape the wildcard (e.g.-e '*/data'
or-e \*/data
), else your shell might expand the pattern. ZFS raw send can be enabled with the-w
flag, in which case compression will be disabled. Resumable zfs send/receive can be enabled with the-r
flag. You can specify a number of retries on connection issues with the--retries
option, and set the retry interval with--retry-interval
. Normally pyznap will not create missing root datasets, but you can set the--dest-auto-create
flag to automatically create it.
-
-
Take snapshots according to policy in default config file:
pyznap snap --take
-
Clean snapshots according to policy in another config file:
pyznap --config /etc/pyznap/data.conf snap --clean
-
Take and clean snapshots according to policy in default config file:
pyznap snap
-
Backup snapshots according to policy in default config file:
pyznap send
-
Backup a single filesystem locally:
pyznap send -s tank/data -d backup/data
-
Send a single filesystem to a remote location, using
pigz
compression:pyznap send -s tank/data -d ssh:20022:root@example.com:backup/data -i /root/.ssh/id_rsa -c pigz
-
Pull a single filesystem from a remote location:
pyznap send -s ssh::root@example.com:tank/data -d backup/data -c lz4
-
Pull a single filesystem from a remote location and send it to another remote location:
pyznap send -s ssh::root@example1.com:tank/data -d ssh::root@example2.com:backup/data -j /root/.ssh/id_rsa_1 -k /root/.ssh/id_rsa_2
-
Backup a single filesystem and exclude some datasets:
pyznap send -s tank -d backup/tank -e '/tank/data*' '/tank/home/user1*' '*/user2/docs'