prometheus-diskspin-exporter

Note: this project was renamed from prometheus-hdparm-exporter to avoid a possible name conflict.

prometheus-diskspin-exporter is a zero-dependency Python program that will export the power state of the SATA disks found on the host and expose them to an HTTP endpoint for consumption by Prometheus.

It monitors if the disks are correctly shut down according to scheduled time in the disk or manual calls to hdparm -y. It calls hdparm -C on every request and parses the status of each SATA disk /dev/sd* reported by lsblk. The program limits queries on SATA disks because NVME SSDs and other disk types will often response with an unknown status to the hdparm -C command.

This way one can monitor disk spin status to be sure that the spin policy works correctly: energy is saved and noise is reduced when not in use, and disks do not spin up and down too often to prevent wear.

Example grafana output graph

The above shows disks in a ZFS pool spinning up after trying to access the filesystem.

Usage

prometheus-diskspin-exporter HOST PORT

Example

$ sudo prometheus-diskspin-exporter localhost 28101 &
$ curl localhost:28101/metrics
hdparm_disk_power_status{disk="/dev/sda",status="standby"} 1 1731616722000
hdparm_disk_power_status{disk="/dev/sdb",status="standby"} 1 1731616722000
hdparm_disk_power_status{disk="/dev/sdc",status="standby"} 1 1731616722000
hdparm_disk_power_status{disk="/dev/sdd",status="active/idle"} 1 1731616722000
127.0.0.1 - - [14/Nov/2024 21:38:42] "GET /metrics HTTP/1.1" 200 236

The last line is the log line coming from the server’s stdout in the background; it is not present in the HTTP response.

Status values:

active/idle means the disk is spinning
standby means the disk is not spinning
unknown means the disk did not report anything useful

Install

The code fits a single file main.py and can easily be distributed on hosts manually:

curl https://raw.githubusercontent.com/gagath/prometheus-diskspin-exporter/refs/heads/main/src/prometheus_diskspin_exporter/main.py \
    -o /opt/prometheus_diskspin_exporter.py
chmod +x /opt/prometheus_diskspin_exporter.py

If you want to install the systemd service:

curl https://raw.githubusercontent.com/gagath/prometheus-diskspin-exporter/refs/heads/main/systemd/prometheus-diskspin-exporter.service \
    -o /etc/systemd/system/prometheus-diskspin-exporter.service

If you want to install the AppArmor configuration:

curl https://raw.githubusercontent.com/gagath/prometheus-diskspin-exporter/refs/heads/main/apparmor.d/opt.prometheus_diskspin_exporter \
    -o /etc/apparmor.d/opt.prometheus_diskspin_exporter

Grafana configuration

The hdparm_disk_power_status metric must be queried as a range table.

Grafana query configuration: choose the "Table" format

You will need to do a bit of transformations of the hdparm_disk_power_status field in Granafa to obtain the State timeline output showed in the introduction:

$Grafana transform data configuration: first grouping to matrix on disk column, then convert field type Time\disk to Time$

Recommended value mappings:

Grafana value mappings: Map the expected hdparm outputs to display appropriately.

Known limitations

The server will reply for any URL, not only /metrics. This is because this is the only thing the server do and there is no point in parsing the query url to match it.

The server will run lsblk and hdparm -C on every request. This may DDoS your machine if an attacker spams the server. Always make sure to allow access to this HTTP server only for authorized users.

License

GPL-3.0-only.