Water, Doors, and Smoke: Sensors Beyond Temperature
Once temperature monitoring is in, most admins consider the physical world handled. The physical world disagrees. It has water: dripping from the AC unit, seeping from the water heater one room over, arriving through the ceiling from the kitchen upstairs. It has doors that get propped open with a chair "for a minute" and stay open all summer. And now and then it has the smell of hot plastic, which every experienced admin can identify from across a building and which no monitoring dashboard has ever mentioned.
None of these are exotic risks. They are slow, quiet, and cheap to detect, and almost nobody detects them, because they sit outside the comfortable world of pings and CPU graphs. Water is the patient one. A drip can run for weeks inside a wall or under a raised floor, choosing its moment. The difference between "we replaced a ceiling tile" and "we replaced the storage array" is usually just who found out first. I have been on both sides of that sentence. The ceiling tile was cheaper.
This guide is about wiring the physical world into the monitoring system you already run. You will learn where water actually shows up (it has favorite spots), and how the two families of leak sensors differ. You will learn what door sensing buys you, and what smoke sensing can and absolutely cannot be. And you will learn about the humble dry contact — the universal adapter that turns almost any physical event into a signal your monitoring system understands. By the end, "the room is flooding" can be an alert with a timestamp instead of a Monday morning discovery.
In this guide
- Where water actually shows up
- Rope sensors vs spot sensors
- Doors: the propped-open problem
- Smoke: awareness, never life safety
- Dry contacts: the universal adapter
- Getting the signals into monitoring
- Alert routing for physical events
- When the water alert fires
- Where to go next
Where water actually shows up
Sensors only help where the water goes, so start with a survey, not a shopping cart. Walk your room slowly and think like a drop of water: it falls from things that contain it, and then it flows downhill. That single sentence predicts nearly every server room flood ever cleaned up. The stops on the walk:
- Under and around the air conditioner. Cooling produces condensation all day, every day, and it is supposed to leave through a small drain line. That drain clogs with algae and dust, and the condensate pan overflows. Then the AC becomes the single most common source of water in equipment rooms — the machine protecting your gear, dripping on it instead. If you place only one sensor in your life, place it here.
- Anything with plumbing nearby. The water heater in the adjacent storage room, the janitor's sink across the hall, the restroom that shares a wall, pipes crossing the ceiling void. Water does not respect room labels.
- The floor's low points. Floors are never truly level. Pour a cup of water in a corner (or just look where the dust stains gather) and you learn where any future leak will pool. Low points are where a sensor detects water from sources you never predicted — the catch-all position.
- The ceiling above the racks. Look up. Stained tiles mean water has traveled through that void before and remembers the route. Anything above you — roof, upstairs kitchen, HVAC on the floor above — can arrive via the ceiling. Where it lands is decided by the void, not by you.
- Under the raised floor, if you have one. A raised floor is a hidden basement where water can spread invisibly across the whole room while the surface stays dry. It is also where the power and network cabling live. Sensors go under there, at the low spots.
Write down what the walk taught you — three to six spots, ranked by how likely and how expensive. Count the neighbors' plumbing as your own, because the wall between you is a delay rather than a barrier. And rank the ceiling honestly. Water that arrives from above aims for the tallest, most expensive rack out of sheer narrative instinct. That list is your sensor placement plan, and it costs nothing but half an hour of thinking like water.
Rope sensors vs spot sensors
Leak sensors come in two families, and the difference is coverage shape. A spot sensor is a small puck or probe pair that alarms when water touches that exact point — a sentry standing on one tile. A rope sensor is a cable that detects moisture anywhere along its length. It is a tripwire you can lay around a perimeter, along a wall, or in a loop under the AC. Neither is better; they answer different questions:
| Spot sensor | Rope sensor | |
|---|---|---|
| Coverage | One precise point | A line or perimeter, meters long |
| Best at | Known targets: drain pans, low points, under a specific unit | Unknown paths: room perimeters, cable trays, under raised floors |
| Location precision | Exact — the alarm names the spot | Somewhere along the rope (some models can report the distance) |
| Installation | Place it and forget it | Routing takes planning; keep it flat on the floor |
| Classic mistake | Guarding the wrong point | Coiled leftover rope in a pile, guarding nothing |
A sensible small-room setup mixes them. It has spot sensors at the AC pan and the known low point. It has one rope run along the wall shared with plumbing or around the racks. Whatever you install, test it the honest way once it is wired into monitoring: touch a wet cloth to the sensor and watch the alert arrive. A leak sensor that has never seen water is a decoration with a cable.
Placement has its own small craft. Sensors and ropes must lie flat on the actual floor — a rope draped across a cable bundle detects rain and nothing else. Tape or clip them in place and label them boldly, because to anyone else a rope sensor looks exactly like a cable that fell down. Helpful people tidy it onto a hook. Expect one education cycle with whoever cleans the building: the first mop bucket or floor-washing machine that meets your rope will page someone. And that is not a false alarm so much as a successful test with community involvement. Learn the cleaning schedule, and treat a wet-floor alert during it with suspicion but never with automation that ignores it. The one real leak that coincides with mopping day would enjoy that loophole far too much.
Northgate Retail had a rope sensor along the back wall of their stockroom server closet, and it worked. It fired on the first Tuesday after installation, when the floor was washed, and again the following Tuesday. On the third Tuesday someone unplugged it from the gateway so the alerts would stop. The gateway had been wired so that an unplugged sensor looked exactly like a dry floor, a choice the section on dry contacts below explains how to avoid. Four months later the water heater on the other side of that wall failed overnight. The rope lay in the water with nothing to say. The review did not blame the person who unplugged it; they had been paged three Tuesdays running for a mop. It blamed the alert rule, which was fair, and the wiring, which was fairer.
Doors: the propped-open problem
The server room door is a boring hinge with two serious jobs. It is part of the cooling envelope — the room is a box the AC can win inside, and only inside. And it is the security boundary between "authorized people" and "everyone who has ever wandered a hallway." A propped-open door quietly fails both jobs at once. The AC now tries to cool the corridor and loses, and temperatures drift up. Meanwhile the room full of your organization's most important equipment is open to anyone carrying confidence and a clipboard.
The fix is a door contact sensor — two small pieces, one on the door and one on the frame, that together report open or closed. It is the same device alarm systems have used forever, and it feeds monitoring beautifully. The design detail that matters is the alert rule: do not alert on the door opening — doors are for opening. And an alert per visit is how the sensor gets muted by Friday. Alert on the door being open too long — longer than any legitimate visit, say tens of minutes. Alert after hours too if opening at 3 a.m. is genuinely unusual for you. The open-too-long alert is precisely the propped-door detector, and it pairs wonderfully with temperature data. The door opens, and intake temps drift up. And your monitoring tells the whole story as one timeline instead of two mysteries.
One honest caveat: if the door keeps getting propped, ask why before escalating. I once found a server room door held open by a fire extinguisher, and the person who put it there had a reason. Almost always someone props a door because the room beyond it is unbearable to work in, or because carrying equipment through a self-locking door alone is a circus act. That is a cooling problem or a process problem wearing a security costume, and a doorstop policy will not fix either. Give people a legitimate answer — fix the heat, add a door holder that alarms politely after a grace period, or station a cart. Then the sensor goes back to catching the genuinely forgotten door, which is the one that matters.
Smoke: awareness, never life safety
First, the honest boundary, stated plainly. The smoke detection that protects human lives is the building fire alarm system, required by codes, professionally installed and inspected. It is not a monitoring project, must never depend on your monitoring server, and is not what this section is about. Nothing you wire into a network dashboard replaces any of it, and any sensor vendor implying otherwise should be shown the door (whose sensor will log the event).
What monitoring can add is awareness for the IT-scale events below the fire alarm's concern. A smoke or particulate sensor in the room, feeding your monitoring system, can notice the early signature of a failing power supply or a cooking capacitor. That is the hot-plastic phase — before there is anything a life-safety system would call a fire. That is a real gift. "Something in the room has started to burn itself out" is exactly the alert you want while the problem is still one device with a fan on it. Many buildings also allow the fire panel itself to expose a simple status output that your monitoring can watch. The output is for awareness, not action. It lets the timeline of any event include it, and lets IT learn that something happened in the building at all. Awareness in, authority out: that is the whole arrangement.
Dry contacts: the universal adapter
Most of these sensors reach your monitoring through one humble mechanism. A dry contact is the simplest interface in the physical world. It has two terminals that are either connected (closed) or not connected (open), like a light switch with no light attached. "Dry" just means the contact itself carries no power of its own — whatever reads it supplies a tiny sensing voltage and asks one question: open or closed?
The reason this matters: an enormous range of equipment speaks dry contact as its native tongue. AC units offer a contact that closes on fault. Generators report running or failed. Leak controllers, door sensors, float switches in sump pits, industrial equipment of every vintage — all of them can raise a hand through two screws and a bit of wire. The dry contact is the universal adapter between the physical world and your monitoring, and it predates every protocol you know by generations.
One convention is worth learning because it decides whether failures hide from you. A circuit can treat "normal" as either open or closed — called normally open and normally closed. Prefer normally closed for anything important, meaning the healthy state is a completed circuit. Then a cut wire, a corroded terminal, or an unplugged sensor breaks the circuit and looks like an alarm — the failure announces itself. With a normally open sensor, a broken wire looks exactly like everything being fine, forever. That is the wiring Northgate Retail had, and an unplugged rope is just a broken wire with a story. Choosing the convention where damage is loud instead of silent is the same fail-safe instinct behind all good monitoring: silence should never be the same signal as safety.
Here is the whole chain in one concrete example. Your AC unit has a pair of terminals labeled as its alarm output, documented to close on fault. Two thin wires run from those terminals to input three on your sensor gateway. In the gateway's configuration you name that input "AC unit fault contact" and declare its normal state. Your monitoring system polls the gateway and now owns a check that has nothing to do with networking and everything to do with Tuesday going well. The day the compressor trips, the contact closes, and the poll notices. The alert says which machine to worry about — often long before the room temperature would have told the same story. Total moving parts: two screws, two wires, one honest label.
Getting the signals into monitoring
Sensors need a translator to reach your dashboards. The usual one is a sensor gateway (often sold as an environmental monitoring unit). It is a small networked box with jacks for temperature and humidity probes, leak ropes, and a row of dry contact inputs. It reads them all and speaks SNMP to your monitoring system. Each sensor becomes a value you poll like any other, with traps for instant notification on change. Give each input a human name in its configuration ("NE corner floor rope", "AC fault contact"), because that name is what appears in the alert. And 3 a.m. is not the time to decode input_4.
You may already own a translator without knowing it. Many UPS management cards accept an environmental accessory with probe and contact inputs. That makes the UPS the natural sensor hub for a small room — one networked device reporting power and climate and leaks. However the signals arrive, two placement rules apply. The gateway plugs into UPS-protected power, because environmental monitoring is most interesting during power trouble. And poll it at the same calm rhythm as other environment data — every minute or so. Enable traps so a wet rope is news in seconds, not at the next poll.
Then monitor the messenger. A dead gateway means every sensor behind it has gone silently blind, which is worse than having no sensors, because you believe you are covered. Alert when the gateway itself stops answering polls, exactly as you would for a switch. "The thing that watches the water has stopped talking" is a real alert with a real response. It is the difference between a sensor system and a sensor-shaped feeling of safety.
Alert routing for physical events
Physical alerts differ from digital ones in a way routing must respect: the response is physical. Nobody can restart a service to dry a floor. Someone with keys, shoes, and proximity has to walk toward the problem. Often the right someone is not IT at all but facilities, the landlord's emergency line, or the HVAC contractor. So the routing rules for water, door, and smoke-awareness alerts deserve their own thought. They go to whoever holds keys and lives nearest. They include the facilities contact from the start. And they never route into a quiet ticket queue to be discovered during business hours. Water at 2 a.m. is exactly as wet as water at 2 p.m., and it uses the extra twelve hours industriously.
The message itself matters more than for any other alert class, because the responder may be half-asleep and unfamiliar with your labels. Compare:
Bad: ALERT env-gw-02 input 4 state=1
Good: WATER DETECTED — Main server room (bldg A, ground floor),
floor rope along north wall, near the AC unit.
Next step: see the water runbook. Facilities on-call:
listed in the on-call sheet.
The good version answers where, what, and what now, in the first two lines. Every physical alert should pass the stranger test: could a capable person who has never seen your room act on this message alone? Write the alert text when you configure the sensor, while you still remember what input 4 means.
Tip: physical sensors deserve scheduled honesty checks. Once in a while, wet the rope with a cloth, prop the door, and confirm each alert arrives, routed to the right people. A sensor is a promise about the future; testing is how you keep it — and a five-minute drill beats discovering a dead gateway on flood day.
When the water alert fires
Detection is half the job; the first minutes of response are the other half, and they go better decided in advance. The sequence that serves most rooms starts with confirm — a second sensor, a camera, or a human look, in that order of speed. Then stop the source if it is stoppable. That is why the runbook must say where the water shutoff valve for your floor is, and who holds the key to reach it. Those details should not be left to anyone's memory. Then protect the gear — move what can be moved, and get cardboard and covers over what cannot. Shut equipment down cleanly if water is approaching it rather than waiting for the decision to be made for you.
And one rule that outranks all equipment: electricity and standing water share a room only on their terms, not yours. Never step into pooled water near powered equipment or power distribution. If the water has reached outlets, power strips, or the bottoms of racks, there are two choices. Cut power to the affected circuits at the breaker — if that can be done safely from dry ground. Or stand back and let facilities and professionals handle it. Every server in that room can be replaced by a purchase order. Write that sentence into the water runbook too. The people most likely to respond at night are exactly the people most likely to wade in to save a switch.
Where to go next
Survey where water would actually go, and cover the top spots with the right sensor shapes. Give the door an open-too-long rule. Add smoke awareness without ever touching life safety's job. And let dry contacts carry everything else into a gateway your monitoring polls. None of this is expensive, and all of it converts silent physical failures into alerts with timestamps and names. The physical world will go on disagreeing with you. The point is to hear about it first, from a sensor rather than a ceiling tile. The same thinking scales down to every wiring closet you own — smaller room, same physics, less supervision. The guide on the closet nobody checks takes it there. The climate guide covers the temperature half in depth. The rest of Power & Environment Monitoring rounds out the picture.
Put this into practice with SysPeep Monitor
Once a sensor gateway or a UPS management card is reporting your physical sensors over SNMP, the rest is ordinary monitoring work. SysPeep Monitor polls those values alongside your network gear and applies your alert rules. It notifies the people you choose the moment a rope gets wet or a door stays open. The free trial installs on a Windows machine in minutes.
Download Free TrialLearn About SysPeep