Systemd Units¶

The systemd probe supervises systemd units on Linux via D-Bus, reporting the active state, sub-state and restart counter per unit. One metric per unit, tagged with systemd.unit.

Linux only. Returns an error on Windows and macOS.

Quick start¶

# probes.d/10-systemd.yaml — each file under probes.d/ is a YAML array of probes
- name: systemd
  type: systemd

All non-transient units are monitored by default; every parameter is optional.

Parameters¶

Parameter	Must set	Default	Description
`units`	No	-	Unit names or shell globs to watch; empty watches every unit of the included types. Example: `nginx.service, ssh*.service`
`include_types`	No	`[service socket timer mount]`	Unit type suffixes to include. Example: `service, timer`
`interval`	No	`30`	Seconds between collections

Without units, every non-transient unit of the included types is watched. Restrict units on a host with many units: each one costs several series per cycle.

Metrics¶

Metric	Unit	Description
`systemd.unit.active_state`	1	1 when the unit active state is `active`, 0 otherwise, tagged with `systemd.unit`
`systemd.unit.sub_state`	1	1 when the unit sub-state is `running` or `listening`, 0 otherwise; the raw sub-state value is available in the `sub_state` tag
`systemd.unit.load_state`	1	1 when the unit load state is `loaded`, 0 otherwise
`systemd.unit.restarts`	{restart}	Number of times the unit has been restarted by systemd; service units only

Operational notes¶

Without units, the probe watches every loaded unit of the included types, except the units systemd creates for each user login and removes at logout (run-user-*.mount, user@*.service, user-runtime-dir@*.service). Services started on demand (packagekit, systemd-oomd) also leave systemd's list when they stop. With a PRTG sensor, which keeps every channel it has seen, name the units to watch in units so the channel list stays stable.
The probe reads from the local system D-Bus socket. No special privileges are needed beyond D-Bus access, which is granted to root by default.
The systemd.unit.type tag carries the unit type suffix (service, socket, mount, timer, …).

Metric reference¶

Every metric this probe can emit. Metric is the OpenTelemetry name the OTLP, Prometheus and Zabbix outputs derive theirs from. Name is what a Nagios check and the API metrics= filter match. PRTG channel is the label PRTG shows, placeholders filled from the series' tags.

Metric	Name	PRTG channel	Unit	Description
`systemd.unit.active_state`	`systemd.unit.active_state`	Systemd {systemd.unit} Active State	#	1 when the unit active state is 'active', 0 otherwise
`systemd.unit.sub_state`	`systemd.unit.sub_state`	Systemd {systemd.unit} Sub State	#	1 when the unit sub state is 'running' or 'listening', 0 otherwise. The sub_state tag carries the raw sub state value.
`systemd.unit.load_state`	`systemd.unit.load_state`	Systemd {systemd.unit} Load State	#	1 when the unit load state is 'loaded', 0 otherwise (not-found, error, masked)
`systemd.unit.restarts`	`systemd.unit.restarts`	Systemd {systemd.unit} Restarts	#	Cumulative restart count for service units (NRestarts D-Bus property). Only emitted for units of type 'service'.