Monitoring HPE Managed PDUs in Zabbix: power, outlets, breakers and rack sensors

A PDU is the least glamorous box in the rack — until a breaker trips and half the servers lose one power supply, or a new server pushes a phase over its limit and the whole rack goes dark. Managed PDUs measure all of this. The data just needs to end up somewhere people actually look.

HPE Managed (Metered and Switched) PDUs expose everything over SNMP, but there was no usable Zabbix template for them, so we wrote one.

Template: custom-zabbix-templates/hpe-pdu — Zabbix 7.4+, SNMPv2c. Tested on HPE Managed PDU P9S18A (230 V, 32 A, 7.4 kVA, 24 outlets, 3 sections).

What it monitors

The template uses the HPE CPQPOWER MIB (pdu3 branch, .1.3.6.1.4.1.232.165.11) and discovers everything automatically.

The PDU as a whole

• input active power (W), apparent power (VA), reactive power (VAR), power factor, frequency;
• total energy consumption (kWh);
• model, serial number, firmware, rating, outlet count, uptime.

Each input phase / section

• voltage, current and rated current, percent load;
• power (W/VA), energy (kWh), power factor.

Each outlet

• name and connector type (C13, C19, Schuko, …);
• current, percent load, power, energy, power factor.

Each breaker group — current, load, power and breaker status.

Environmental probes connected to the PDU — temperature, humidity and contact/door sensors, each with probe status.

Alerts

All thresholds are macros. The defaults match a 32 A, 230 V European PDU — adjust them for 16 A models or other voltages.

Setup

1. Enable SNMPv2c on the PDU in its web interface and set a community string (not public). Restrict SNMP access to the Zabbix server/proxy if the PDU allows it.

2. Check connectivity from the Zabbix server or proxy:

   snmpwalk -v2c -c <community> <pdu-ip> .1.3.6.1.4.1.232.165.11
   

3. Import the template (Data collection → Templates → Import, hpe_pdu.yaml).

4. Create a host for the PDU with an SNMP interface pointing to its management IP, and link HPE Managed PDU by SNMP.

5. Set the macros: {$SNMP_COMMUNITY}, and the thresholds that differ from the defaults, e.g. for a 16 A PDU {$INPUT.CURRENT.WARN}=12.8 and {$INPUT.CURRENT.CRIT}=14.4.

Polling: 60 s for live values and 300 s for energy counters ({$UPDATEINT}, {$UPDATEINT.SLOW}); history 90 days, trends a year.

What to do with the data

• Capacity planning. Phase load trends tell you how much headroom each rack really has — instead of adding up nameplate wattages, which are always too pessimistic.
• Phase balancing. On three-section PDUs, one overloaded section trips while the others are half empty. Per-phase graphs make imbalance obvious; move a few power cords before it becomes an incident.
• Redundancy check. With two PDUs per rack (A/B feed), both should carry a similar load. If one PDU suddenly drops to zero on several outlets, a server PSU or a cable has failed — and the rack is running without redundancy.
• Power cost per server. Name the outlets in the PDU after the devices plugged into them; per-outlet energy (kWh) then gives you real consumption per server — useful for chargeback or for deciding which old hardware to retire.
• Physical security. A door sensor on the rack plus the contact trigger gives you an audit trail of who opened the rack and when.

Summary

• One SNMP template for HPE Managed PDUs: input, phases, outlets, breakers, sensors.
• Discovery does the work — outlets and probes appear automatically.
• Alerts for overload, tripped breakers, voltage, temperature, humidity and open doors.
• Tune thresholds for your PDU’s rating.

Need monitoring for your server room or rack infrastructure? Let’s talk.