SNMP temperature OIDs, by vendor

# every sensor value on the device
snmpwalk -v2c -c public 10.0.0.10 1.3.6.1.2.1.99.1.1.1.4

# the sensor names, so you can tell which index is the inlet
snmpwalk -v2c -c public 10.0.0.10 1.3.6.1.2.1.47.1.1.1.1.2

snmpwalk -v3 -l authPriv -u monitor \
  -a SHA -A '<auth-pass>' -x AES -X '<priv-pass>' \
  10.0.0.10 1.3.6.1.2.1.99.1.1.1.4

Standard — ENTITY-SENSOR-MIB (RFC 3433)

Much modern Cisco, Juniper and Arista gear, and plenty of servers. Try this first.

Object	OID	What it gives you
entPhySensorValue	1.3.6.1.2.1.99.1.1.1.4	The reading itself. Raw integer — must be scaled.
entPhySensorType	1.3.6.1.2.1.99.1.1.1.1	Unit of the reading. 8 = celsius. Check it rather than assuming.
entPhySensorScale	1.3.6.1.2.1.99.1.1.1.2	Power-of-ten multiplier. 9 = units (10⁰), 8 = milli, 10 = kilo.
entPhySensorPrecision	1.3.6.1.2.1.99.1.1.1.3	Decimal places implied. 1 means divide by 10.
entPhySensorOperStatus	1.3.6.1.2.1.99.1.1.1.5	1 = ok, 2 = unavailable, 3 = nonoperational. Discard anything not ok.
entPhysicalDescr	1.3.6.1.2.1.47.1.1.1.1.2	ENTITY-MIB. Human names — how you find which index is the inlet.

Scaling: value × 10^(scale − 9) ÷ 10^precision. In practice most gear reports Celsius with precision 0 or 1.

Walk entPhysicalDescr alongside entPhySensorValue and match the indexes — the descriptions tell you which sensor is inlet, exhaust, CPU or PSU. If the walk returns nothing, fall back to the vendor profile below.

Dell — PowerEdge via iDRAC

iDRAC 7, 8 and 9. This is the profile where the scaling trap bites.

Object	OID	What it gives you
temperatureProbeReading	1.3.6.1.4.1.674.10892.5.4.700.20.1.6	The reading, in tenths of a degree.
temperatureProbeLocationName	1.3.6.1.4.1.674.10892.5.4.700.20.1.8	Probe name, e.g. System Board Inlet Temp.
temperatureProbeStatus	1.3.6.1.4.1.674.10892.5.4.700.20.1.5	3 = ok. Discard readings that aren't.

Scaling: divide by 10. A returned 275 is 27.5 °C. This is the single most common mistake made with iDRAC.

Index 1 is normally System Board Inlet Temp — that is the one you want for room or rack monitoring. Higher indexes are CPU and exhaust probes, which run much hotter and will give you a misleading picture of your cooling.

Step-by-step for Dell: enabling SNMP on the card, identifying the inlet probe, verifying the reading and setting a threshold that means something.

HPE — ProLiant via iLO

iLO 4, 5 and 6, with the SNMP agent enabled — it is off by default on some firmware.

Object	OID	What it gives you
cpqHeTemperatureCelsius	1.3.6.1.4.1.232.6.2.6.8.1.4	Current temperature, whole degrees Celsius.
cpqHeTemperatureLocale	1.3.6.1.4.1.232.6.2.6.8.1.3	Where the sensor is: 3 = system, 4 = system board, 6 = CPU, 7 = memory, 11 = ambient/intake.
cpqHeTemperatureThreshold	1.3.6.1.4.1.232.6.2.6.8.1.5	The threshold HPE itself considers critical for that sensor.

Scaling: none — whole degrees Celsius.

Walk cpqHeTemperatureLocale first and find the entry with value 11 (ambient). That is your intake reading; everything else is internal and will alarm for reasons unrelated to the room.

Step-by-step for HPE: enabling the iLO SNMP agent, finding the ambient sensor, the two-part index that trips up every first attempt, and why HPE's own threshold is not an alert level.

Cisco — IOS, IOS-XE, NX-OS

CISCO-ENVMON-MIB, for switches and routers with environmental sensors.

Object	OID	What it gives you
ciscoEnvMonTemperatureStatusValue	1.3.6.1.4.1.9.9.13.1.3.1.3	Current reading, whole degrees Celsius.
ciscoEnvMonTemperatureStatusDescr	1.3.6.1.4.1.9.9.13.1.3.1.2	Sensor description — inlet vs outlet vs CPU.
ciscoEnvMonTemperatureState	1.3.6.1.4.1.9.9.13.1.3.1.6	The device's own verdict: 1 normal, 2 warning, 3 critical, 4 shutdown, 5 notPresent, 6 notFunctioning.

Scaling: none — whole degrees Celsius.

Newer IOS-XE also answers on ENTITY-SENSOR-MIB; prefer that when it responds, since it states scale and precision. ciscoEnvMonTemperatureState is a useful second opinion, but don't rely on it alone — vendors set those thresholds conservatively for the chassis, not for your room.

Juniper — Junos (MX, EX, QFX, SRX)

Object	OID	What it gives you
jnxOperatingTemp	1.3.6.1.4.1.2636.3.1.13.1.7	Temperature of an operating component, Celsius.
jnxOperatingDescr	1.3.6.1.4.1.2636.3.1.13.1.5	Which component the index refers to — Routing Engine, FPC, chassis, PSU.

Scaling: none — whole degrees Celsius.

Indexed per operating component, so one device returns many rows. Walk jnxOperatingDescr first and pick the chassis or intake entries; the Routing Engine always reads hotter and is not a proxy for room temperature.

APC / Schneider Electric — NMC and rack PDUs

PowerNet-MIB. Network Management Card (AP9630/9631/9640/9641) with an environmental probe such as the AP9335T, and rack PDUs. This is where the units trap bites hardest.

Object	OID	What it gives you
iemStatusProbeCurrentTemp	1.3.6.1.4.1.318.1.1.10.2.3.2.1.4	Probe temperature.
iemStatusProbeName	1.3.6.1.4.1.318.1.1.10.2.3.2.1.2	Probe name as configured on the card.
iemStatusProbeCurrentHumid	1.3.6.1.4.1.318.1.1.10.2.3.2.1.6	Relative humidity, percent.
iemStatusProbeTempUnits	1.3.6.1.4.1.318.1.1.10.2.3.2.1.5	1 = celsius, 2 = fahrenheit. Check this.

Units: whatever the card is configured for. An NMC set to Fahrenheit returns 78 for a perfectly normal 25.5 °C room, and a naive threshold of 32 will never fire. Poll iemStatusProbeTempUnits alongside the reading, or set the card to Celsius and write that down somewhere.

Rack PDUs on newer firmware expose sensors under the rPDU2Sensor tree instead. If the OIDs above return nothing, walk 1.3.6.1.4.1.318.1.1.26.

Step-by-step for APC: which card and probe you need, handling the unit correctly, humidity, and where rack PDUs hide their sensors.