Winter in the Server Room: The Annual Check for Cooling, Power, UPS and Alerting

There is exactly one moment every year where a mid-market server room is really put to the test — and it is not summer. It is the three weeks between Christmas and New Year. The office is empty, the second-line admin is on vacation, the facilities manager is skiing in South Tyrol, and the one remaining number in the alert chain lands in a mailbox because the colleague on duty would rather end the year in Thailand. That is exactly the window where the previous twelve months of server-room care either hold up — or where you spend January 2nd flying home from a ruined ski trip.

The annual winter check in the server room is therefore not just another maintenance appointment. It is a structured template covering cooling, power, UPS, fire protection and alerting, worked through before year-end and documented before the first out-of-office emails roll in. This post summarises which items actually matter from our practice as a mid-market IT house, where we regularly find gaps in customer projects, and what the checklist looks like that we use to carry our own customers through the winter.

Why the winter check differs from summer maintenance

In summer, everybody expects the air conditioning to work hard. Thirty-two degrees outside, twenty-four inside — that gets noticed. In winter the unpleasant things happen quietly. The outdoor condensers ice up, the pressure vessel compensates for a while until the high-pressure switch finally trips. The UPS is rarely tested under full load because everything looks stable in normal operation. The smoke detectors were last pressed years ago. The server room door contact has not been properly wired back into the alarm panel since the last office renovation.

There is also a meteorological angle: outage frequency on the German grid is measurably higher in winter. Storm gusts, freezing rain, trees on overhead lines — the SAIDI statistics show winter as the most stressful season for the medium-voltage grid. So if your UPS has to switch to battery operation more often in winter, it is worth making sure beforehand that it actually can.

And finally humidity. In a heated building, relative humidity in January and February regularly falls below twenty-five percent. That is not only uncomfortable for staff, it also significantly raises the risk of electrostatic discharge — especially when hardware is being touched during maintenance.

HVAC service before the cold season

The first and most important building block of the year-end check is the cooling system. We are not just talking about the indoor units inside the server room, but above all about the outdoor units that have to work under significantly more adverse conditions from December onwards.

These items belong in the service, ideally performed by a certified refrigeration specialist and not by the facilities caretaker:

• Refrigerant level and leak testing — legally required once a year above 5 tonnes CO2 equivalent, worth doing on smaller systems anyway
• Condenser cleaning — leaves, pollen and late-autumn dust in the finned heat exchanger
• Condensate drain inspection — drain lines running outside sometimes freeze shut in winter, backing water up into the indoor unit
• Fan rotation and bearing check — a dead outdoor fan is obvious in summer, but can go unnoticed for weeks in winter
• Defrost cycle verification — heat pumps and reversible splits must defrost reliably at outdoor temperatures below 5 degrees
• Redundancy test: if you have two AC units, the second must carry full server load alone — switch the primary unit off for 30 minutes and watch the room temperature

If your HVAC is not redundantly dimensioned, the winter check is the right moment to address it. We have documented the sizing fundamentals in our article Server Room: Planning Power and Cooling for the Mid-Market.

A practical note on free cooling: if the AC system has an economiser mode using cold outside air, check the switchover setpoints for winter. We regularly see installations where free cooling kicks in at too-high outdoor temperatures and then drives indoor humidity far below twenty percent — which punishes electronic components and spinning disks.

UPS battery test and year-end maintenance

The UPS is the area where we see unpleasant surprises most often in customer projects. The unit stands there, the status LED is green, the last self-test passed — and when the first real outage hits in January, it delivers seven minutes instead of the projected sixty.

The year-end UPS check goes significantly deeper than the automatic self-test:

Full-load battery test under real load. Not the built-in dummy test, but a real discharge against actual server load. The approach: document load and time, force the UPS into battery mode via the management interface until either remaining capacity drops to 30 percent or 15 minutes are reached, then switch back to mains and compare the time actually delivered against the projected runtime. If the real time is below 70 percent of nominal, batteries are due.

Log the ambient temperature. Lead-acid battery life halves per 10 degrees above 20. If your UPS sits in a constant 26-degree room, the projected 5 years are more like 3 years. For the full sizing logic, see our article Right-Sizing UPS for the SMB Server Room.

Check firmware status. UPS devices have been hit by several critical security advisories in recent years (SmartUPS SNMP, APC Powerchute, Eaton Intelligent Power Manager). A firmware update is not something you do on a Tuesday in March — but the year-end window is the right moment.

Test the shutdown automation. Pull the mains plug, start the stopwatch, document the order in which VMs, hypervisors and storage come down. If TrueNAS shuts down before its iSCSI clients, you have a problem — and you only find out in a real test. In our Proxmox and TrueNAS managed service, that test is a fixed year-end item.

Document the maintenance bypass. When battery replacement is due in January, the bypass switch has to be not only present but labelled — ideally with the date of the last inspection.

Smoke detectors, door contacts and physical sensors

The server room is well-equipped technically in most mid-market companies and under-equipped sensorically. Smoke detectors, door contacts, water sensors and temperature probes are often installed for years, but nobody checks whether they still feed meaningfully into the alert chain.

These items belong on the year-end list:

• Smoke detector functional test — press the test button, see the signal in the monitoring, document it
• Smoke detector age — lithium smoke detectors have a 10-year manufacturer warranty, after which replacement is mandatory
• Door contact — does it correctly trigger the alarm panel? Does it send a signal into the IT monitoring when the door is open?
• Temperature and humidity probes — calibration against a reference device; we regularly find probes drifting 5 degrees
• Water sensor in the raised floor — functional test with a damp cloth
• Lightning protection inspection — legally every four years, but the certificate belongs in the yearly audit folder

With temperature and humidity probes it is worth revisiting the alert thresholds. We recommend yellow alarm (SMS, email) at 26 degrees, red alarm (phone call, escalation) at 29, and a hard shutdown at 32. Humidity: alarm below 25 percent and above 70 percent. Those values can shift with the hardware mix, but they make a good starting point.

Alerting chain and escalation over the holidays

Technology is one half, process the other. The best sensor is worthless if the alert hits a mailbox or lands in an email address nobody is reading. The winter check is the right moment to review the alert chain and adjust it for Christmas absences.

In practice:

1. First stage — SMS or push to the primary on-call admin, with acknowledgement expected within 10 minutes
2. Second stage — on missing acknowledgement, escalate to a second admin, ideally not in the same household (not both on the same vacation)
3. Third stage — management or external service provider (such as us as an external IT house) is activated, with a clear mandate to act
4. Fourth stage — on-site contact, usually a facility manager or security service with access rights and a written emergency plan for server-room access

In our managed service offering, we take on the second or third stage for many customers. The important point: before year-end we need to know what we are allowed to do on site and which hands-on actions are authorised — a clear emergency plan, not just a phone number.

Test the chain once. A real test alarm in November, with a stopwatch, with documentation. Who was unreachable first, where the acknowledgement stalled, how long the end-to-end runtime took — those numbers surface the gaps that would otherwise appear in January under production conditions.

Monitoring and logging: what should be re-baselined at year-end

The final element of our winter check is the monitoring itself. We see two recurring themes: monitoring gaps on newly added systems (a switch has arrived during the year but no SNMP target was configured), and sprawling alert configurations where every second alarm is a false positive. If half of all alerts are noise, the admin will ignore the real incident too.

Year-end monitoring checklist:

• Completeness — every system in the server room should be captured in at least one monitoring dimension (ping, SNMP, API)
• Alert density — how many alerts came in during the past month? More than ten per day is too much
• Threshold review — values untouched for two years rarely match the current load profile
• Backup of the monitoring configuration — an export of your Zabbix, PRTG or Checkmk config belongs in the central backup
• Credential rotation — SNMP communities, API tokens, admin passwords, if unchanged for over a year

A practical tip: year-end is the ideal moment to visualise the past twelve months of monitoring data in one view. CPU, network, storage and temperature curves over 12 months show trends that get lost in day-to-day operations — and they are the best input for next year’s budget planning.

FAQ on the server-room winter check

When should the annual check be performed?

Ideally in November, by early December at the latest. That leaves enough time to fix issues before the holidays. If you schedule the check for the first week of January, you risk discovering problems only after the critical period is over.

Which parts of the check can internal IT handle, and where do you need external specialists?

Monitoring, UPS configuration and door contact tests can typically be fully covered by internal IT. For refrigeration work, lightning protection inspection and electrical certification you need certified contractors — for liability reasons, not only technical ones.

How often should a real UPS load test be run?

At least once a year under real load. Ideally four times, i.e. quarterly, with a lighter test (switchover time and battery voltage, for example) and a full-load test at year-end.

What does annual maintenance of cooling and UPS cost roughly?

That depends heavily on system size and location. For a typical mid-market server room with two AC units and a 5 kVA UPS, the pure maintenance cost per year sits in the low four-figure range. Battery replacement, if due, comes on top. For a concrete number against your setup, get in touch.

Which documents should the winter check leave behind?

A protocol per topic (cooling, UPS, sensors, alert chain, monitoring), signed by the technician performing the work and by the IT owner, with date and defect list. This documentation is not only operationally important, it is also increasingly required by cyber insurance and audit regimes.

What happens if we skip the annual check?

Not necessarily anything terrible straight away — but the probability that an incident goes undetected or hits at the worst possible time of year goes up noticeably. Beyond that, cyber insurers increasingly argue the demonstrability of operational discipline during claim adjustment, including documented maintenance of ancillary infrastructure.

Summary

The annual winter check in the server room is the cheapest form of operational insurance we know of in mid-market IT. A structured walk-through of cooling, UPS, sensors, alert chain and monitoring, with clean documentation, prevents in our experience the majority of Christmas and New Year incidents that have crossed our desks over ten years of practice.

DATAZONE supports mid-market companies from Bavaria and beyond with exactly this year-end maintenance — from the technical check of ancillary systems through the review of the alert chain to taking on the second or third escalation stage over the holidays. If you want your server room cleanly walked through before December 20th, get in touch. We keep capacity for this slot reserved in the final weeks of every year.