Ansible Role zfs-auto-snapshot

This is an ansible role for installing and configuring zfs-auto-snapshot.

Requirements

Cron installed
Python (for ansible)
ZFS installed and zfs command in $PATH
Root (or sudo) access
Internet access for ZFS hosts (download script)
The ZFS datasets need to exist, they won't get created
Should run on any system that runs POSIX shell. Though tested only on Debian.

Usage

Example usage:

- hosts: zfs-servers
  vars:
    zfs_datasets:
      tank/data/home:
        frequent: true
        daily: true
        monthly: true
      tank/data/public:
        hourly: true
        daily: true
  tasks:
    - import_role:
        name: ccremer.zfs-auto-snapshot

Parameters

Advanced usage (with default values):

- hosts: zfs-servers
  vars:
    zfs_datasets:
      # No worries, the actual default is empty. This example dataset shows the default properties of each dataset if not defined.
      tank/example/dataset:
        enabled: true
        frequent: false
        hourly: false
        daily: false
        weekly: false
        monthly: false
    # How often the 'frequent' snapshots occur. Integers from 1 to 30
    zfs_keep_frequent_interval: 15

    # Count how many snapshots we can have before deleting.
    zfs_keep_frequent: 4
    zfs_keep_hourly: 24
    zfs_keep_daily: 31
    zfs_keep_weekly: 8
    zfs_keep_monthly: 12

    # Should we always download the most recent script with each run?
    autosnap_update: false
    # The arguments to append to the script.
    autosnap_args: "--quiet --syslog --default-exclude"
  tasks:
    - import_role:
        name: ccremer.zfs-auto-snapshot

Gotchas

The script provided by zfs-auto-snapshot by default snapshots ALL datasets it can find. Personally, I find this rather dangerous in terms of space consumption. So this playbook inverts the default behavior by adding --default-exclude to the arguments. This makes it necessary to explicitly define and enable the datasets in your playbook. You can change this back by removing the --default-exclude flag if you like (in which case you would define the datasets that should not get snapshot).

Contributions

Bugs

Please identify whether the behavior is caused by zfs-auto-snapshot itself before opening an issue here.

Feature requests

Hmm, maybe. If I have time and motivation. Consider forking/+PR if you really need something more.

Alternatives

zfs-auto-snapshot is simple and doesn't rely on additional dependencies other than cron. If you prefer more control, check out zfsnap or zfs-snap-manager.

ccremer/ansible-zfs-auto-snapshot