Remote site UPS monitoring over SNMP comes down to a few values from the standard UPS-MIB (RFC 1628): the output source, which shows when the load is on battery, the battery status, the estimated minutes remaining and the input voltage. Poll them every minute, alarm on the first three, and you know which sites need a visit before they go dark.
I work on field connectivity for unmanned sites. The UPS-MIB is a standards-track MIB, so the list below works on any UPS network card that implements it.
Which UPS values should you monitor over SNMP?
Watch the output source, the battery status and the estimated minutes remaining first; they answer whether the site is on battery and how long it has. The UPS-MIB in RFC 1628 defines all of them:
| Object | What it says | Alarm on |
|---|---|---|
| upsOutputSource | Where the output power comes from: normal, bypass, battery and others | battery(5) |
| upsBatteryStatus | unknown, batteryNormal, batteryLow or batteryDepleted | batteryLow(3), batteryDepleted(4) |
| upsEstimatedMinutesRemaining | Estimated minutes to depletion under the present load | Below your travel time to the site |
| upsInputVoltage | Input voltage, RMS volts | Outside the band your mains should hold |
| upsTestResultsSummary | Result of the last self-test | doneError(3) |
batteryLow means the estimated run-time is at or below upsConfigLowBattTime, a threshold in minutes set on the UPS itself, so read it once per site. batteryDepleted means the UPS could not carry the present load if mains were lost, which is worth a visit even while mains is fine.
Should you use SNMP traps or polling for a UPS?
Use polling as the baseline and traps as a faster extra. A trap is a single unconfirmed datagram: RFC 3416 states there is no confirmation for the SNMPv2 trap, while the InformRequest is confirmed, though still without a guarantee of delivery.
Polling also catches the case traps never can: a site that has gone completely silent. RFC 3417 suggests UDP port 161 for agents and port 162 for notification receivers, so both directions have to pass the site router; keep that path on a management segment, as described in IoT network segmentation for plants. The UPS-MIB’s upsTrapOnBattery is resent every minute until the UPS turns off or leaves battery.
Hermetry connects field devices to live dashboards, rules and alerts: IoT telemetry and SCADA in one platform. Its SNMP templates reference includes a UPS-MIB (RFC 1628) template for UPSs with a standard network card, polled every 60 seconds with numeric OIDs, so no MIB file is needed. It stores values such as output.source, battery.status and battery.minutes_remaining, and the page lists alarm conditions such as output.source == “battery”.
What is the difference between Gauge32 and Counter32?
A Gauge32 can go up and down and is read as it is; a Counter32 only goes up and only means something as a difference between two polls. RFC 2578 defines a Counter32 as wrapping to zero after 4,294,967,295 and says a single counter value has, in general, no information content.
In the UPS-MIB, upsAlarmsPresent is a Gauge32: the number of active alarms right now. upsInputLineBads is a Counter32 that goes up by one each time the input goes out of tolerance. A rising line-bads count at a site that rarely runs on battery points at the mains supply, not the batteries.
How do you estimate UPS battery runtime remotely?
Compare upsEstimatedMinutesRemaining with the time a technician needs to reach the site. The MIB defines it as an estimate under the present load, so it changes when the load changes.
- Keep the drive time to each site in its record, and alarm when the UPS is on battery and the minutes remaining fall below that drive time plus a margin.
- Trend the minutes remaining after each outage; a site whose figure keeps shrinking needs new batteries.
None of this replaces the site’s own protection and safe shutdown logic. For how sites with many power assets are watched together, see UPS, generator and climate monitoring at unmanned sites and monitoring for energy sites.
Frequently asked questions
Is SNMPv2c good enough for UPS monitoring at remote sites?
SNMPv1 and SNMPv2c authenticate with a community string sent as plain text, and RFC 3410 calls them fundamentally insecure. Use SNMPv3 where the UPS card supports it, keep the card on a management segment, and never expose it to the internet.
How often should you poll a UPS over SNMP?
Once a minute suits most remote sites. It shows a switch to battery within about a minute, and a missed poll tells you quickly that a site has gone silent. Polling faster adds traffic on the site link without changing what a technician does.
Why does the UPS show minutes remaining while it is on mains?
RFC 1628 defines upsEstimatedMinutesRemaining as the time to depletion under the present load if utility power is off, or if it were lost and stayed off. The value is a forecast even on mains, and it is the number to compare with drive times before an outage.

