By the SNMP Monitoring team · Reviewed July 2026
Water and powered electronics is the one combination that goes from fine to catastrophic in minutes. A CRAC condensate line backs up, a chilled-water fitting weeps, a pipe in the ceiling void lets go over a weekend — and by the time someone walks in, there's standing water under the raised floor and gear is shorting out. Leak detection only helps if it pages you immediately, because the value of the alert decays by the minute. SNMP brings the leak signal into the same monitoring plane as everything else, so a wet-floor sensor can wake the on-call the instant it trips instead of being discovered Monday morning.
This is the how-to. Parent: environmental monitoring. Siblings: smoke & fire, temperature & humidity.
Spot vs rope leak sensors
Two physical designs, and picking the right one is about coverage:
- Spot (point) sensor — a small probe that detects water at one specific location. Cheap, simple, perfect for a known drip point: a CRAC drip pan, under a specific coupling, at the low point of the floor.
- Rope (cable) sensor — a length of water-sensing cable that triggers anywhere along its run. It gives you perimeter or area coverage — snake it around the room edge, along a pipe route, or the full length of a row of racks.
The rule of thumb: spot sensors guard the places water is likely to appear; rope sensors guard the places water would travel to. Most server rooms use both — spot sensors on the obvious drip points, a rope perimeter to catch anything that migrates across the floor.
| Type | Coverage | Best for |
|---|---|---|
| Spot (point) | One specific location | Known drip points — CRAC pans, couplings, floor low points |
| Rope (cable) | Anywhere along its length | Perimeters, pipe runs, whole rows — where water travels |
How leak state is exposed in SNMP
A leak sensor is a dry-contact device: physically, it's just a switch that closes (or opens) when water bridges it. That contact wires into an SNMP environmental gateway, and the gateway publishes the state as an integer enum in its vendor enterprise MIB under 1.3.6.1.4.1 — typically one value for normal/dry and another for alarm/wet.
Two things follow from that. First, there is no averaging or trending a leak — it's boolean, and the only question is which integer means "wet." Second, the polarity is vendor-specific: on one gateway 1 might mean normal and 2 alarm; on another it's inverted. You must read the vendor MIB to learn the mapping, because getting it backwards means your most important alert is silently disabled. Exact OIDs are vendor-specific — don't invent them; describe against the MIB.
Reading the sensor
Poll the gateway's leak object with a simple GET (the arc is a vendor placeholder):
snmpget -v2c -c <RO_string> <gateway> 1.3.6.1.4.1.<vendor>.sensors.leak.1
enterprises.<vendor>.sensors.leak.1 = INTEGER: 1
Now the critical step: look up what 1 means in the vendor MIB. If the MIB defines 1 = normal(1) and 2 = alarm(2), this reading is dry and healthy. If it's the reverse, 1 is an active leak and you're in an incident. Never assume — load the MIB so the value resolves to a name like normal or waterDetected instead of a bare integer. <RO_string> and the <vendor> arc are placeholders; keep live community strings out of committed scripts.
Where to place sensors
Detection is only as good as placement. A practical placement order:
- CRAC/CRAH drip pans and condensate lines — the single most common source of server-room water.
- Under the raised floor at the low points — water flows downhill; put sensors where it will pool.
- Along chilled-water pipe runs and fittings — rope sensor tracing the pipe route catches a weeping joint anywhere.
- Room perimeter and doorways — to catch water migrating in from outside the room.
- Under and behind racks with liquid nearby — anywhere a leak would reach expensive gear first.
The guiding idea: put spot sensors where water starts and rope sensors where water travels, and always cover the lowest points of the floor, since that's where it ends up.
Critical alerting
A leak is a P1. Treat it that way:
- Route it to page immediately — phone/push, not just a dashboard tile or an email that waits for morning.
- Pair polling with a trap so the gateway pushes the alarm the instant the contact trips, rather than waiting for the next poll cycle. See traps vs polling.
- If you must poll, poll fast — a leak found on a 15-minute cycle is a leak found late.
- Alert on sensor/gateway health too — a gateway that stopped answering can't report a leak, and that silence is the dangerous case.
The reason to combine a trap with polling is durability: the trap gives you instant notification, and the poll confirms the state and catches the alarm even if a single trap packet is lost. Route, escalate and de-duplicate through alerts.
Reference
This is the leak-detection how-to. The environmental section of the full sensor index lists related objects at all sensors. Siblings: smoke & fire, temperature & humidity. Definitions at glossary.
Frequently asked questions
Can SNMP detect water leaks?
Yes. Leak detectors are dry-contact sensors wired to an SNMP environmental gateway, which reports the state as an integer enum in its vendor MIB under 1.3.6.1.4.1 — one value normal, the other alarm. You poll (or receive a trap from) the gateway and alert on the alarm state as critical.
What's the difference between a spot and a rope leak sensor?
A spot sensor detects water at one point — ideal for a known drip location like a CRAC pan. A rope (cable) sensor triggers anywhere along its length, giving perimeter or area coverage for pipe runs and room edges. Most rooms use both: spot sensors where water starts, rope sensors where it travels.
How do I get an instant leak alert?
Pair polling with an SNMP trap so the gateway pushes the alarm the moment the contact trips, instead of waiting for the next poll. Route it as a P1 to a paging channel, and poll fast as a backstop in case a trap is lost. Also monitor the gateway's own reachability so a dead probe doesn't hide a leak.
Where should leak sensors go?
Start with the most common source — CRAC drip pans and condensate lines — then the low points under the raised floor where water pools, along chilled-water pipe runs, at the room perimeter, and under racks near any liquid. Spot sensors guard likely sources; a rope sensor covers where water would spread.
Key takeaways
- Leak sensors are dry-contact devices reported as an integer enum in a vendor gateway MIB under
1.3.6.1.4.1. - Spot sensors guard known drip points; rope sensors cover perimeters and pipe runs — most rooms use both.
- Read the vendor MIB for polarity — never guess which integer means "wet," or the alert is silently dead.
- A leak is a P1: page immediately, combine a trap with a fast poll for durability.
- Monitor the gateway's own health — a silent probe can't report a leak.
- Full index: all sensors; an off-box external check via external monitoring still pages you if a room-wide incident hits.