Server Room Monitoring for a Mid-Sized Company: Temperature, Power and the Alarms That Matter
November 11, 2026 · 8 min read · Etzal Group Team
The air conditioner in a trading company's server room trips on a Saturday afternoon. The room is a converted storeroom with no windows. By midnight it is above 45 degrees. Servers shut themselves down one by one. One storage unit does not shut down cleanly. On Monday morning, staff arrive to find no network, no file server and a day of data recovery ahead. The air conditioner had a fault light, glowing unseen behind a closed door. Server room monitoring would have sent a message at 15:20 on Saturday, when the temperature first passed 28.
What can go wrong in a small room
- Cooling fails. The most common serious event. A sealed room with running equipment heats up within an hour or two.
- Power fails, and the uninterruptible power supply runs out before anyone knows.
- Water arrives: a leaking air conditioner, a pipe above, a blocked drain.
- The UPS battery has aged and gives minutes instead of the expected half hour.
- Someone leaves the door open, or enters when they should not.
- Smoke or overheating equipment.
Each gives warning. The question is whether anyone receives it.
Six things to measure
1. Temperature at the rack inlet. At the front of the racks, where equipment takes in air. Not on the wall, and not at the back where hot air leaves.
2. Humidity. Too high risks condensation and corrosion. Too low increases static. A broad band of roughly 20 to 80 percent relative humidity is workable for most equipment. A sharp change often indicates a cooling fault or an open door.
3. Mains power. Is utility power present? A simple sensor on a wall socket not backed by the UPS tells you the moment it is lost.
4. UPS status. On battery or on mains, load percentage, estimated runtime remaining, battery health. Most UPS units can report this over a network card or a serial connection.
5. Water on the floor. A leak sensor under the air conditioner, along the wall with pipes, and at the lowest point of the floor.
6. Door state. Open or closed, with the time. Useful for security and for explaining a temperature rise.
Worth adding where the risk justifies it:
- Smoke detection linked to the monitoring system, in addition to the building's own.
- Air conditioner status: running, fault, supply air temperature.
- Power per circuit, to see load creeping toward a breaker's limit.
- A camera triggered by door opening.
Where to put temperature sensors
Placement determines whether the reading means anything.
- Front of rack, upper third. Hot air rises and recirculates over the top, so the highest equipment gets the warmest intake.
- Front of rack, middle, on the racks that hold critical equipment.
- One at the cooling unit's supply, to see what the air conditioner is delivering.
- One at the back or ceiling, for the return air.
For a room with two or three racks, four to six sensors give a useful picture.
The difference between supply and inlet temperature shows how well cool air reaches the equipment. The difference between front and back shows the load.
Thresholds that work
Two levels, with a delay to avoid alarms from brief fluctuations.
| Measurement | Warning | Critical |
|---|---|---|
| Rack inlet temperature | Above 27 for 10 minutes | Above 32 for 5 minutes |
| Rate of temperature rise | More than 3 degrees in 15 minutes | |
| Humidity | Outside 30 to 70 percent for 30 minutes | Outside 20 to 80 percent |
| Mains power | Lost for more than 1 minute | |
| UPS on battery | Immediately | Runtime below 15 minutes |
| UPS battery health | Replace-battery flag | |
| Water detected | Immediately | |
| Door open | Open longer than 10 minutes | Opened out of hours |
Adjust these to your equipment and room. Two points matter more than the exact numbers:
- Alert on the rate of rise. A jump of several degrees in a quarter of an hour means cooling has failed, well before the absolute limit is reached. It buys the most time.
- Set the warning well below the danger level, so there is time to act.
Alerts must not depend on what they monitor
This is the design point most often missed.
When power fails or the room overheats, the switches, the firewall and the internet line may go down. If the monitoring system sends alerts through them, it falls silent at the worst moment.
- Its own power. A battery that keeps the monitor alive for hours after the UPS is exhausted.
- Its own way out. A cellular connection, independent of the office network.
- Watched from outside. A cloud service that expects a regular heartbeat and raises an alarm when the heartbeat stops. Silence must itself be an alert.
Who gets told, and how
- Critical alerts to at least two named people, by a channel that wakes them: a phone call or a loud app notification, not email.
- Acknowledgement required, with escalation to the next person if nobody responds in ten minutes.
- Warnings to the IT team in working hours.
- A message that says what to do: "CRITICAL: Server room temperature 33 and rising, cooling suspected failed. Check AC unit. Building security: 021-xxx."
Include someone who can physically reach the room. For many companies that is building security or a facilities contact, who should have access, a simple instruction sheet and the authority to act.
The response plan
Write it before the first alarm, on one page, and keep a printed copy by the door.
Temperature critical:
- Check the air conditioner. Reset it if tripped.
- If it will not restart: open the door, place portable fans, bring a portable cooler if one is kept on site.
- Call the air conditioning contractor. Number on the sheet.
- If the temperature keeps rising past a set point, shut down non-essential equipment in the listed order.
- At the next set point, shut down everything in a controlled way. A clean shutdown is far better than a thermal one.
Mains lost:
- Confirm whether it is the building or the utility.
- Note the UPS runtime.
- If no generator: start the controlled shutdown sequence when runtime reaches the agreed threshold.
- Remember that cooling usually stops with mains power, even while the UPS keeps servers running. The room heats faster than expected.
Water detected:
- Find the source. Close the valve if it is a pipe.
- Protect equipment. Cut power to affected circuits if water is near them.
- Call the building.
The shutdown order, written in advance, is the most valuable part. Nobody should be deciding at 02:00 which server matters least.
Reduce the risk as well as watching it
Monitoring tells you. These make the event less likely or less damaging.
- Redundant cooling. Two units that alternate, each able to carry the load.
- Automatic restart of the air conditioner after a power cut. Many units stay off until someone presses a button.
- Regular air conditioner service, and a contractor with an agreed response time.
- UPS battery replacement on schedule. Batteries typically last a few years, less in a warm room.
- A runtime test twice a year, so the stated minutes are real.
- Automatic graceful shutdown of servers when the UPS runtime falls below a threshold.
- Nothing stored in the room that blocks airflow.
- No water pipes overhead, or a drip tray with a sensor where they cannot be moved.
Test it
- Monthly: check that every sensor is reporting.
- Quarterly: trigger each alarm deliberately. Warm a temperature sensor, wet a leak sensor, unplug the mains sensor. Confirm the right people receive the message.
- Twice a year: a UPS runtime test, and a walk through the response plan with the people named in it.
- After any change in equipment or contacts: update the thresholds and the sheet.
An alarm that has never been tested is a hope.
Common mistakes
- One temperature sensor on the wall.
- Alerts by email only.
- Alerts that travel through the network being monitored.
- No alert when the monitor itself goes silent.
- Thresholds set at the danger point, leaving no time to act.
- One person on the alert list, who is on holiday.
- No one with physical access in the response chain.
- No written shutdown order.
- An air conditioner that does not restart after a power cut.
- A UPS battery nobody has tested.
We design and install environmental monitoring for server rooms and plant rooms as part of IoT and Control Systems. If your server room has a thermostat and nothing else, contact us.
Summary
Measure temperature at the rack inlet, humidity, mains power, UPS status, water and door state. Set warning and critical thresholds with delays, and alert on a fast temperature rise. Give the monitor its own battery and cellular path, and treat its silence as an alarm. Send critical alerts by phone to at least two people, including someone who can reach the room. Write the response and the shutdown order in advance, reduce the risk with redundant cooling and tested batteries, and test every alarm each quarter.
Frequently asked questions
What temperature should a server room be?
Industry guidance from ASHRAE recommends air entering the equipment at roughly 18 to 27 degrees Celsius for typical servers. What matters is the temperature at the front of the racks, where equipment draws air in, not the reading on the wall thermostat. Check your own equipment's stated operating range as well.
Why does monitoring need to work when the network is down?
Because the events you most need to hear about, such as a power failure or overheating, often take the network down with them. If alerts depend on the same switches and internet line that have just failed, they never arrive. The monitoring device needs its own battery and its own path out, typically a cellular connection.
Is a small server room worth monitoring if most systems are in the cloud?
Usually yes. Even with applications in the cloud, the room typically holds the network core, firewalls, local storage, CCTV recorders and telephone equipment. If that room overheats or loses power, the office cannot reach the cloud. A basic set of sensors costs far less than a day without the network.
Related service
WhatsApp AI Chatbot
We build, train, and run a WhatsApp AI agent that answers customers 24/7, qualifies leads, and follows up.
