71 lines
2.3 KiB
Markdown
71 lines
2.3 KiB
Markdown
# ZFS
|
|
|
|
Ansible community support for ZFS is limited to create filesystems, volumes and snapshots. There is no support for managing zpools, so here it is.
|
|
|
|
## Inventory
|
|
|
|
Here is an example inventory file you can use with this role:
|
|
|
|
```yaml
|
|
zfs_pools:
|
|
- name: peace
|
|
type: raidz1
|
|
devices:
|
|
- ata-SOME-DISK-LABEL-1
|
|
- ata-SOME-DISK-LABEL-2
|
|
options:
|
|
ashift: 12
|
|
root: /mnt/peace
|
|
state: present
|
|
```
|
|
|
|
And you will get raid1 zpool peace with two disks, with 12 ashift.
|
|
|
|
You can use a variety of options, see the [zpoolprops(7)](https://openzfs.github.io/openzfs-docs/man/master/7/zpoolprops.7.html) man page.
|
|
|
|
And for your zfs filesystems:
|
|
|
|
```yaml
|
|
zfs_datasets:
|
|
- name: peace/pictures
|
|
state: present
|
|
- name: peace/movies
|
|
state: present
|
|
extra_zfs_properties:
|
|
mountpoint: /mnt/peace/movies
|
|
quota: 500G
|
|
```
|
|
|
|
## Prometheus metrics
|
|
|
|
The role deploys an hourly systemd timer (`zfs-metrics.timer`) running
|
|
`/usr/local/bin/zfs-metrics.sh`, which writes pool and dataset capacity to
|
|
node_exporter's textfile collector directory (`zfs_metrics_dir`). The
|
|
`node_exporter` role must be deployed on the host; the script exits silently
|
|
otherwise.
|
|
|
|
| Metric | Labels | Description |
|
|
| --- | --- | --- |
|
|
| `zfs_pool_size_bytes` | `pool` | Total pool size |
|
|
| `zfs_pool_alloc_bytes` | `pool` | Allocated bytes |
|
|
| `zfs_pool_free_bytes` | `pool` | Free bytes |
|
|
| `zfs_pool_capacity_percent` | `pool` | Used capacity |
|
|
| `zfs_dataset_used_bytes` | `dataset` | Used bytes (incl. snapshots and children) |
|
|
| `zfs_dataset_refer_bytes` | `dataset` | Own referenced bytes |
|
|
| `zfs_dataset_quota_bytes` | `dataset` | Quota (0 = none) |
|
|
|
|
These feed the **ZFS - Pools & Datasets** Grafana dashboard. node_exporter's
|
|
built-in `--collector.zfs` exposes ARC, I/O and pool state only, not capacity.
|
|
|
|
`zfs_pool_capacity_percent` also drives Grafana capacity alerts (80/90/95%
|
|
tiers) to ntfy, provisioned by the `grafana` role — see `roles/grafana/README.md`.
|
|
|
|
Tunables (`defaults/main.yml`): `zfs_metrics_enabled`, `zfs_metrics_dir`,
|
|
`zfs_metrics_script_path`, `zfs_metrics_interval`.
|
|
|
|
## References
|
|
|
|
- https://docs.ansible.com/ansible/latest/collections/community/general/zfs_module.html
|
|
- https://github.com/mrlesmithjr/ansible-zfs/blob/master/tasks/manage_zfs.yml
|
|
- https://wiki.archlinux.org/title/ZFS
|