IPMI / BMC Sensors¶

The ipmi probe reads hardware sensor data from the local machine's Baseboard Management Controller via ipmitool, reporting temperatures, fan speeds, voltages and power supply status.

Linux only. Requires ipmitool and the OpenIPMI kernel driver (ipmi_si).

Quick start¶

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

For local sensors, no parameters are needed. The probe calls ipmitool sdr against the local BMC.

Parameters¶

Parameter	Must set	Default	Description
`mode`	No	`local`	local reads the host's own BMC; remote polls a BMC over LAN. One of `local`, `remote`
`remote`	If `mode` is `remote`	-	Remote BMC access, used with mode remote
`remote.host`	No	-	BMC address or hostname; required with mode remote
`remote.username`	No	-	IPMI user
`remote.password`	No	-	IPMI user's password. A secret: reference it with `${secret:…}`, `${env:…}` or `${file:…}` rather than writing it in the file
`remote.interface`	No	`lanplus`	ipmitool interface; lanplus for IPMI 2.0, lan for IPMI 1.5
`sensors`	No	-	Sensor selection
`sensors.include_types`	No	-	Only these sensor types; empty means all. Example: `Temperature, Fan`
`sensors.exclude_names`	No	-	Regular expressions of sensor names to skip
`ipmitool_path`	No	`ipmitool`	Path of the ipmitool binary when it is not on the PATH. Example: `/usr/bin/ipmitool`
`interval`	No	`60`	Seconds between collections
`exec_timeout`	No	`10`	Seconds an ipmitool run may take before it is killed

To poll a BMC over the network instead of the host's own, set mode: remote and give the BMC under the remote block. Sensor filters live under the sensors block:

# probes.d/10-ipmi.yaml
- name: ipmi
  type: ipmi
  params:
    mode: remote
    remote:
      host: 10.0.0.50
      username: monitor
      password: ${secret:ipmi.password}   # OS secret store; inline plaintext is auto-sealed on install
      interface: lanplus                  # lan for an IPMI 1.5 BMC
    sensors:
      include_types: [Temperature, Fan]
      exclude_names: ["^PSU2.*"]

Metrics¶

Metric	Unit	Description
`hardware.temperature`	Cel	Temperature per BMC sensor, tagged with `hardware.component`
`hardware.fan.speed`	RPM	Fan speed per sensor
`hardware.voltage`	V	Voltage per sensor
`hw.power`	W	Power per sensor (a supply, or the machine's consumption)
`senhub.hardware.current`	A	Current per sensor, typically per power supply
`hw.status`	#	Sensor status: 1 = ok or non-critical, 0 = critical or non-recoverable

Operational notes¶

The agent must run as root on Linux for ipmitool to access the /dev/ipmi0 device.
Load the OpenIPMI driver before starting the agent: modprobe ipmi_si && modprobe ipmi_devintf.
Sensor names (the hardware.component tag) come directly from ipmitool sdr output and vary by hardware vendor.
Sensors that share a name (Dell calls every CPU temperature Temp) are told apart by their IPMI entity: Temp (CPU 1), Temp (CPU 2). These series also carry the hardware.entity and hardware.sensor_number tags. A sensor whose name is unique keeps it unchanged.
A power supply reports 1 when present and redundant, 0 on a critical status, "Redundancy Lost" or a failure; an absent supply ("Device Absent") emits nothing.

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
`hw.temperature`	`hardware.temperature`	Temp	Degrees Celsius	Temperature reported by the BMC sensor (degrees Celsius)
`hw.fan.speed`	`hardware.fan.speed`	Fan Speed	RPM	Fan speed in rotations per minute as reported by the BMC
`hw.voltage`	`hardware.voltage`	Voltage	Volts	Voltage reported by the BMC sensor (volts)
`hw.power`	`hardware.power`	Power	Watts	Power reported by the BMC sensor (watts): a supply's output or input, or the machine's consumption
`senhub.hardware.current`	`senhub.hardware.current`	Current	Amps	Current reported by the BMC sensor (amperes), typically per power supply
`hw.status`	`hardware.power_supply.status`	PSU Status	#	Power supply status: 1 = ok (present, redundant, or a healthy reading), 0 = fault (critical status, redundancy lost or degraded). An absent supply or one the BMC has no reading for emits no value
`hw.status`	`hardware.sensor.status`	Sensor	#	Generic BMC sensor status: 1 = ok or non-critical, 0 = critical or non-recoverable. A sensor with no reading (ns) emits no value
`senhub.ipmi.up`	`senhub.ipmi.up`	IPMI Reachable	#	1 when ipmitool ran successfully and the BMC responded, 0 otherwise