Table of Contents
- Why a Cooling Maintenance SLA Needs More Than a Response Time
- The Data Centre Cooling Maintenance Checklist to Attach to Your SLA
- Setting Cooling System SLA Response Times and Restoration Targets
- Building a Chiller Maintenance Schedule Into the Agreement
- Data Centre Maintenance Reporting: Evidence the SLA Was Met
- Planned Maintenance Without Downtime: Windows, Redundancy and Change Control
- Frequently Asked Questions
Last Updated: October 18, 2026
Why a Cooling Maintenance SLA Needs More Than a Response Time
A data centre cooling maintenance SLA is a written agreement that sets out what cooling work gets done, how fast a provider must respond when something fails, and how both sides prove the work happened.
A response time tells you when someone answers the phone.
At Treske, we design, supply and install power and cooling infrastructure, then maintain it. We have seen what happens when an SLA measures the wrong thing. The room drifts.
Below, we break down what a cooling SLA should actually contain, from the maintenance checklist to the reporting that proves compliance.
The Data Centre Cooling Maintenance Checklist to Attach to Your SLA
The checklist is the backbone of the agreement. Without it, the SLA has no substance.
A practical data centre cooling maintenance checklist should cover:
- Filter changes and coil cleaning
- Refrigerant and pipework checks
- Condensate drain and pump inspection
- Fan and belt inspection
That list is the floor, not the ceiling. A checklist that survives an audit names the task, the equipment, the cadence, the acceptance criterion and the record produced. A line that says "inspect filters" is not measurable.
What the Checklist Should Cover by Equipment Type
Different cooling equipment needs different tasks. A CRAC unit needs filter and humidifier checks. A chiller needs refrigerant and water treatment checks. In-row cooling needs airflow and condensate checks.
| Equipment | Core Tasks | Typical Cadence | Evidence Produced |
|---|---|---|---|
| CRAC/CRAH units | Filters, coils, humidifier, condensate drain, controls | Monthly to quarterly | Filter pressure log, coil condition photo, controls test sheet |
| In-row cooling | Airflow, condensate, fan check, setpoint verification | Quarterly | Airflow reading, fan current draw, setpoint record |
| Chillers | Refrigerant and leak check, water treatment, strainer clean, compressor test | Quarterly to annually | Leak test record, water-treatment report, compressor log |
| Cooling towers and dry coolers | Basin clean, fill and drift eliminator check, fan and belt inspection, water treatment | Monthly to quarterly | Basin condition photo, belt tension record, chemical dosing log |
| Pumps and pipework | Seal and bearing check, valve exercise, insulation condition | Quarterly to annually | Vibration or temperature reading, valve exercise sheet |
| Environmental monitoring | Sensor calibration, alarm test, failover test | Annually (calibration), quarterly (alarm test) | Calibration certificate, alarm test log |
| Liquid-cooling loops | Leak detection, fluid quality and filtration, connector inspection, load test | Per manufacturer guidance, typically quarterly to annually | Fluid sample result, leak-detection test record, connector torque log |
Cadence is an example only. Check your own contract and manufacturer guidance, a site with high particulate load or a coastal environment may need shorter intervals than a clean inland site.
Turning the Checklist Into an Enforceable Clause
A checklist only changes behaviour when the contract attaches consequences to it. Three clauses do most of the work:
- Task completion as a service level. State that scheduled tasks must be completed within the agreed window, and that a missed task counts as a service-level breach unless a documented exclusion applies.
- Evidence per task. Require a record for each task, a reading, a photo, a signed sheet or a sensor export, retained for an agreed period and available on request.
- Deferral approval. Any task deferred must be approved in writing by the customer, with a revised date, so deferrals cannot quietly become permanent.
For teams that need ongoing support, our preventative maintenance service covers scheduled servicing across cooling and power equipment.
Setting Cooling System SLA Response Times and Restoration Targets
Cooling system SLA response times need to be split into separate measures. Most contracts blur them into one, which lets a provider meet the letter of the deal while missing the point. The fix is a severity model with a measurable target at each stage, plus rules for how each target is measured.
Response, Restoration and Resolution Are Not the Same Thing
Response time is how long until a qualified person acknowledges the fault and starts work. Restoration time is how long until cooling capacity is back. Resolution time is how long until the root cause is fixed.
Define all three. Set separate targets for each. A fast response with a slow restoration is not good service.
A Severity Model for Cooling Incidents
Cooling incidents are not all equal. A single fan failure in an N+1 room is a different event from a chiller trip that leaves the room on one unit. Agree severity tiers in the contract, with a target for each stage. The ranges below are typical examples only, set your own with your provider and check them against your redundancy design.
| Severity | Example trigger | Acknowledgement | Attendance | Workaround | Restoration to design capacity |
|---|---|---|---|---|---|
| Critical | Loss of cooling capacity with no redundancy remaining; room temperature rising above the agreed band | Within minutes, 24/7 | Within hours | Temporary cooling deployed or load reduced within hours | Within an agreed short window, typically same day |
| High | Loss of one unit in an N+1 configuration; redundancy reduced but room stable | Within an hour, 24/7 | Same day | Redundancy restored or temporary unit in place within an agreed window | Within one to two business days |
| Medium | Degraded performance, alarm on a non-critical component, no capacity loss | Next business day | Within an agreed window | Not usually required | Within an agreed planned window |
| Low | Cosmetic or advisory fault, scheduled task follow-up | Within the reporting cycle | Scheduled visit | Not required | Next planned maintenance window |
Making the Targets Measurable
A target is only enforceable if both sides agree how it is measured. Put these rules in the contract:
- Clock start. Define when the clock starts, typically when the alarm is received by the provider's monitored service desk, not when an engineer is dispatched.
- Clock stop. Define what counts as restoration, for example, cooling capacity returned to the agreed design level and verified by monitoring, not merely an engineer on site.
- Measurement window. State whether targets apply 24/7 or only in business hours, and whether they are measured per incident or as a monthly average. Per-incident targets are stricter and easier to audit.
- Exclusions. List what does not count against the provider, customer-caused delays, access refused, parts on extended lead time agreed in advance, and force majeure. Exclusions should be narrow and documented, not a catch-all.
Reporting and Evidence
Targets need a paper trail. Require a monthly report showing each incident, its severity, the acknowledgement, attendance, workaround and restoration times, and whether each target was met. Back it with monitoring data, alarm timestamps, temperature and humidity trends, and unit status logs, so the report can be checked against actual conditions rather than accepted on trust. Retain the records for an agreed period so an audit can reconstruct any incident.
Roles, Escalation and Change Control
A severity model only works if everyone knows who does what. Name the customer contact who can authorise access, load reduction or emergency spend, and the provider's escalation path from service desk to account manager to technical lead. Agree how changes to the SLA, new equipment, revised targets, added sites, are approved, so the agreement stays current as the room evolves.
Building a Chiller Maintenance Schedule Into the Agreement
The chiller maintenance schedule belongs in the contract, not in a separate document nobody reads. Chillers carry the bulk of the cooling load, so their upkeep drives reliability.
A chiller maintenance schedule should set out:
- Quarterly refrigerant and leak checks
- Water treatment and strainer cleaning
- Compressor and control testing
- Annual overhaul and efficiency review
- Load testing after major service
Put the cadence in the SLA as a table. Add a clause that missed tasks count as a breach. That single clause changes behaviour.
For teams that need ongoing support, our preventative maintenance service covers scheduled servicing across cooling and power equipment.
Data Centre Maintenance Reporting: Evidence the SLA Was Met
Data centre maintenance reporting is how you prove the SLA was honoured. Without evidence, the agreement is just words. Require a written report on a fixed cycle.
What a Monthly Report Should Contain
A monthly report should include:
- Tasks completed against the schedule
- Any missed or deferred tasks, with reasons
- Faults logged, response and restoration times
- Temperature and humidity trends
- Alarm events and how they were handled
Pair the report with live monitoring. Tools like EkkoSense data centre optimisation give real-time visibility of temperature, humidity and cooling performance, so the report can be checked against actual conditions rather than a summary.
Planned Maintenance Without Downtime: Windows, Redundancy and Change Control
Planned maintenance should not mean planned outage. With the right redundancy and change control, you can service cooling equipment while the room stays live.
Three things make this work:
- Redundancy. N+1 or better cooling lets you take one unit offline safely.
- Maintenance windows. Agree them in advance. Off-peak is standard.
- Change control. Every task follows a written change process with rollback steps.
Test failover before the window, not during it. Confirm the remaining units can carry the load.
For facilities that need extra assurance, our site acceptance testing work verifies that cooling systems perform as designed before and after major changes. And if you run precision units in smaller rooms, the Vertiv Liebert DM range is built for 24/7 operation with precise temperature and humidity control.
A cooling SLA is only as strong as the tasks and evidence behind it.
Treske designs, supplies and installs cooling, power and rack infrastructure, then keeps it running with scheduled maintenance and monitoring. We back our work with extended warranty options and ongoing support.
Visit us today to review your cooling maintenance SLA and build one that actually protects uptime.
Frequently Asked Questions
What is the difference between response time and resolution time in a cooling maintenance SLA?
Response time is how long the provider takes to acknowledge an incident and begin work, usually measured from the moment a fault is logged. Resolution or restoration time is how long until cooling capacity is back to the agreed level. A contract can promise a fast response and still leave you waiting days for a fix, so insist on both. Ask for restoration targets stated per severity level, and confirm whether the clock runs during business hours only or around the clock.
How often should data centre cooling systems be serviced?
Most cooling maintenance agreements split tasks across monthly, quarterly and annual cadences. Monthly work typically covers filter checks, airflow inspection and alarm review. Quarterly visits add coil cleaning, condensate drain checks and refrigerant pressure readings. Annual servicing covers compressor checks, electrical torque testing and full performance verification. Chillers usually need their own schedule, often with a pre-summer inspection. Treat these as example ranges only and confirm the actual cadence in your own contract.
How can cooling equipment be maintained without disrupting data centre operations?
Planned maintenance works around redundancy. The provider isolates one cooling unit at a time while the remaining units carry the load, and the work is scheduled in an agreed maintenance window with a change record raised beforehand. Before any isolation, check that the remaining cooling capacity covers current rack density and that temperature monitoring is live. Agree in the SLA what happens if redundancy is already reduced, since that is when a routine service becomes a risk.
What information should be included in a cooling maintenance report?
A useful data centre maintenance reporting pack shows the tasks completed against the checklist, any faults found and how they were closed out, temperature and humidity readings from the monitoring system, and SLA performance against response and restoration targets. It should also flag trends, such as rising return air temperatures or a chiller working harder than the season warrants. Ask for the raw monitoring export, not just a summary, so you can verify the claims independently.