Prometheus metrics exporter for ZFS.
It's much the same data you get from zpool status and zpool iostat, but in the form of Prometheus metrics. Run
curl -s http://localhost:9254/metrics
to see them.
Sample dashboard is available at grafana.net.
Currently doesn't handle pool changes gracefully, you'll have to kill and restart it if you create/import or destroy/export any pools.
libzfs is not a stable or official interface, so this could break with any new ZFS release.
Requires root privileges on Linux. For the security conscious, run it with -web.listen-address=localhost:9254.
If you're not running prometheus on the same host, either use my exporter-proxy or write a little cronjob that does
curl -s http://localhost:9254/metrics > mydir/zfs.tmp && mv mydir/zfs.tmp mydir/zfs.prom
Configure node_exporter with -collector.textfile.directory=mydir and it will
publish the stats to allow remote scraping.
https://github.com/eliothedeman/zfs_exporter
Same idea, but implemented by parsing the output of zpool using github.com/mistifyio/go-zfs.
https://github.com/eripa/prometheus-zfs
Ditto but without the dependency on go-zfs.
PR for node_exporter to add support for ZFS metrics. Currently also relies on shelling out, which is apparently not allowed in node_exporter, so not clear where it's going.
To build and run zfs-exporter in a virtual environment, you need to install VirtualBox, Vagrant, and Ansible 2.2+. Then
platform=jessie # or fedora25
sudo ansible-galaxy install gantsign.golang
cd zfs-vagrant-ansible/
vagrant up $platform
cd ..
ANSIBLE_HOST_KEY_CHECKING=False ansible-playbook -i zfs-vagrant-ansible/.vagrant/provisioners/ansible/inventory/vagrant_ansible_inventory playbook-zfs-exporter.yml -l $platform
