Liquid Cooling Maintenance Checklist: Daily, Monthly and Annual Tasks for CDUs, Cold Plates and Manifolds

Most liquid cooling problems don't start with a failure. They start with drift. A filter loads up a little more each week, the corrosion inhibitor slowly depletes, a quick disconnect seal wears with every mating cycle, and the makeup volume creeps up month after month. None of it shows up until chip temperatures start climbing, and by then you're usually planning downtime.

With AI racks, the stakes are higher. In a technical bulletin on liquid cooling, ASHRAE warned that at extreme chip powers, a cooling outage can have catastrophic consequences. That's why liquid cooling maintenance can't be run-to-failure. The slow risks need to become scheduled checks that are recorded and traceable.

This liquid cooling maintenance checklist is written for preventive maintenance on direct-to-chip liquid cooling in data centers. It covers the secondary loop, or technology cooling system (TCS): the CDU, manifolds, hoses and quick disconnects, cold plates, and the coolant itself. If you're looking for how to clean a PC water-cooling loop or an AIO cooler, there's a short answer in the FAQ.

Liquid Cooling Maintenance Checklist at a Glance

A liquid cooling maintenance checklist, or PM schedule, groups tasks by interval:

• Daily: Review CDU alarms, temperatures, flow, pressure drop and leak sensors.
• Weekly: Walk the loop for drips and log filter ΔP.
• Monthly: Test the coolant.
• Quarterly: Inspect quick disconnects, hoses, valves and pumps.
• Annually: Run lab analysis, calibrate sensors and test failover.

What a Liquid Cooling Maintenance Program Covers

Preventive maintenance for direct-to-chip liquid cooling means inspecting the CDU, manifolds, quick disconnects, cold plates and coolant on a fixed schedule, and comparing every reading against a baseline recorded at commissioning.

A direct-to-chip system typically has two loops. The primary side is the facility water system (FWS), which connects to cooling towers, chillers or dry coolers. The secondary side is the TCS, which runs from the coolant distribution unit (CDU) through the manifolds, hoses and quick disconnects, across the cold plates inside each server, and back to the CDU. Our liquid cooling glossary defines both loops.

This checklist covers the TCS, the loop that sits right next to your IT equipment. The FWS is usually maintained by the facilities team as part of the mechanical plant, so it's out of scope here.

Before you build the checklist, settle three things:

• Ownership: Spell out which parts of the loop the data center operator, the server OEM and the liquid cooling vendor are each responsible for, and who gets the call when something goes wrong.
• Authority: Server OEM and CDU manufacturer manuals come first. The intervals and thresholds in this article reflect common practice; your equipment documentation and warranty terms take precedence.
• Records: Log every reading in a maintenance log or your DCIM system. Liquid cooling maintenance pays off through trends. A single reading tells you very little.

Set a Baseline Before the First Inspection

Maintenance is a comparison against "normal," so start by capturing a baseline at handover:

• Supply and return temperature, total flow and pump speed for each CDU
• Flow and pressure drop for each manifold branch
• Initial differential pressure across clean filters
• Coolant pH, conductivity, glycol concentration and appearance, plus a lab report
• Expansion tank pressure and the system's initial fill volume

The coolant baseline matters most. Field guidance drawn from ASHRAE TC 9.9's bulletin on TCS coolant integrity recommends sending samples to a lab when the loop is first filled, whenever a large amount of makeup fluid is added, and at regular intervals after that. Keep the fluid spec the loop was designed around in your handover documents, so every later test has something to be measured against.

Flushing, filling and air purging belong to commissioning, which this article doesn't cover. For the overall commissioning sequence, see our liquid cooling commissioning guide.

Daily and Weekly Checks

Almost none of this requires downtime. It's mostly a matter of reading your monitoring data:

• Alarm log: Review CDU and DCIM alarms from the past 24 hours, including any that cleared on their own. An alarm that keeps appearing and clearing is often the first sign of trouble.
• Supply temperature: Compare the actual supply temperature with the setpoint. A growing gap points to lost heat exchanger capacity, a problem on the facility water side, or a heat load that has outgrown the CDU.
• Flow and pressure drop trends: If flow holds steady while pressure drop rises, the usual causes are filter loading or early channel fouling. Rule out a change in coolant temperature or glycol concentration first, since both change viscosity.
• Pump speed: On a CDU that controls to a flow setpoint, pump speed that keeps rising at the same load means system resistance is increasing.
• Leak sensors: Confirm every sensor is online and none is offline or bypassed.
• Expansion tank pressure and makeup: A sealed loop should need almost no makeup fluid under normal conditions.

Once a week, walk the floor. Look for wet spots or crystallized residue around manifolds, quick disconnects and hoses, listen for unusual noise from the CDU, and confirm every valve is in the right position. Log filter differential pressure at the same time, so you can see how fast each filter is loading.

Monthly Checks

• Field coolant sampling: Measure pH, conductivity and glycol concentration (a refractometer works for glycol), and check color and clarity. Sample more often in the first few weeks after startup, then drop to monthly once readings stabilize. Our direct-to-chip coolant guide explains what each parameter means and how to correct it.
• Filter replacement planning: Review the weekly ΔP readings against the clean-filter baseline, and schedule a change before any filter reaches the manufacturer's replacement limit.
• Makeup volume reconciliation: Compare this month's makeup volume with last month's, after subtracting fluid drawn for samples or drained during service. Makeup fluid you still can't account for usually means a slow leak somewhere, such as a weeping quick disconnect or a seal that's starting to seep.
• Lead/standby pump rotation: On redundant pump sets, rotate the lead and standby pumps as the manufacturer recommends, so the standby pump isn't sitting idle until the moment you need it.

Quarterly Checks

• Quick disconnects: Check seals for damage or deformation, make sure couplings connect smoothly, and look for dripping after disconnect. Hoses hanging off a quick disconnect put side load on it and can distort the seal, so confirm the hoses are supported. For specs and selection, see our liquid cooling quick disconnect guide.
• Hoses: Look for aging, bulging or abrasion, and make sure no hose is bent tighter than its minimum bend radius.
• Valves: Exercise the isolation valves on the manifolds and CDU to confirm they'll close and open when you need them, since valves that never move can seize. Closing a branch valve cuts cooling to that branch, so do this one valve at a time in a planned window, on branches whose servers are idle or shut down.
• Pumps: Listen for abnormal noise, measure vibration if you can, and compare with last quarter.
• Leak detection functional test: Trigger the sensors using the manufacturer's method and confirm the alarm reaches your monitoring system. If a leak alarm triggers automatic isolation, follow the manufacturer's test procedure so the test doesn't shut off cooling to live servers.
• Electrical and controls: Check terminal connections, control cabinet cooling and communications.

Annual Checks

• Full lab coolant analysis: Beyond the field tests, check inhibitor levels, metal ions (copper, aluminum, iron and others), microbiological activity and particulates. Rising metal ions usually mean corrosion is already underway.
• Sensor calibration: Calibrate flow, pressure, temperature and conductivity sensors. Sensor drift undermines every trend you're relying on.
• Redundancy and isolation testing: In a controlled maintenance window, test pump failover and automatic isolation after a leak, and confirm flow and supply temperature stay within limits during the switchover.
• Heat exchanger performance: Calculate the CDU approach temperature (TCS supply temperature minus the facility water temperature entering the CDU) and compare it with the baseline at a similar load. If the approach has widened while load and facility water flow are unchanged, heat exchanger fouling is the most likely cause.
• Alarm thresholds and firmware: Confirm alarm thresholds still fit the current rack load, and update firmware as the manufacturer recommends.
• Wear parts: Replace seals, pumps and other life-limited parts according to the manufacturer's manual.

Component-by-Component Liquid Cooling Maintenance Checklist

Cold plates themselves are maintained mostly through the fluid. Their channels are narrow: OCP's pre-commissioning cleaning guidance notes that current single-phase cold plates typically use flow channels of around 100 microns. That makes clean coolant and good filtration the most effective cold plate maintenance there is.

When a Check Fails: Escalation Paths

Part of a checklist's job is to route problems to the right process early:

• Separate trends from alarms: A reading that crosses an alarm limit needs immediate action. A reading that's still in range but keeps moving in one direction needs an investigation.
• Rising pressure drop or chip temperature: Follow the sequence in our cold plate cooling troubleshooting guide: review the data, then check settings, valves and trapped air, and pull a cold plate only as a last resort.
• Coolant drift: Follow the coolant guide. The first fix is usually to boost inhibitors or adjust concentration; a full drain and refill is for severe contamination.
• Drips or a leak alarm: Follow the steps in what happens if liquid cooling leaks: isolate first, then find the source.

Servicing Without Taking the Rack Down

Most liquid cooling maintenance can be done with the rack still running, as long as the loop was designed for it:

• Branch isolation: Close a single branch's isolation valves so only one node or one rack is affected.
• Hot swap: On a CDU with redundant pumps and duplex filters, you can replace a pump or filter element while the other set carries the load.
• Quick disconnect handling: Use non-spill (dry-break) quick disconnects, confirm the branch is isolated before you disconnect, and check for residual fluid afterward.
• Top off with the same fluid: Two coolants both labeled "PG25" can carry different additive packages. Before topping off, confirm the brand, concentration and formulation match what's already in the loop.

In-rack and in-row CDUs also service differently. In-rack units mean more units to maintain but lower risk per task, while in-row units mean fewer units but every service window requires confirming the redundant units have picked up the load. Our in-rack vs. in-row CDU comparison covers this in detail.

Your filtration approach affects routine workload too. OCP's PG25 guidelines compare sidestream and full-flow filtration, and OCP's modular TCS white paper includes field experience with 25 µm filtration in the TCS. Finer filtration protects cold plates better but means more frequent element changes, so plan filtration and maintenance together at the design stage.

Design Choices That Make Liquid Cooling Easier to Maintain

Many maintenance headaches are locked in when the hardware is selected. As a cold plate and CDU manufacturer, APALTEK designs for serviceability from the start:

• Redundant pumps: APALTEK's in-house developed and manufactured CDUs use redundant pumps, typically N+1, so the system keeps running while one pump is serviced. See the full range of AIDC thermal management products, including the In-Rack CDU 150 kW and the In-Row CDU 1.5 MW.
• Leak detection with automatic isolation: Our CDUs detect leaks and isolate automatically, cutting the time it takes to respond.
• Pre-installed quick disconnects with assembly leak testing: Cold plates and manifolds can ship with quick disconnects already installed, and the whole assembly is leak tested, which cuts down the number of joints that have to be made in the field.
• Helium leak testing on every product: Every APALTEK liquid cooling product is helium leak tested before it ships. Our helium leak testing guide explains the methods and specs.
• Testing on your coolant: Cold plate thermal tests run on PG25 by default and can be run on water or DI water on request, so test conditions match your site.

See APALTEK's testing and validation capabilities for more.

FAQ

How often should liquid cooling preventive maintenance be scheduled?

Review monitoring data daily, and schedule on-site checks in monthly, quarterly and annual tiers. The exact intervals depend on system size, load and manufacturer requirements, and the server OEM's and CDU manufacturer's manuals and warranty terms take precedence. Check more often in the first weeks to months after startup.

Do cold plates need maintenance?

Cold plates have no moving parts and don't need routine teardown. Monitor them through branch pressure drop and chip temperature. Pull and clean a cold plate only after pressure drop has risen steadily and other causes have been ruled out. Clean coolant and good filtration are the best protection.

Should CDU filters be changed on a schedule or by pressure drop?

Change them primarily by pressure drop, with a time-based limit as a backstop. Differential pressure tells you how loaded the element actually is, which is more accurate than a fixed interval. Watch closely right after startup, when installation debris loads filters faster.

Can liquid cooling maintenance be done without shutting down servers?

Mostly, yes. Monitoring reviews, sampling and walkdowns don't affect operation. On CDUs with redundancy, filter and pump service can be done under load. For quick disconnect or cold plate work on a single node, branch isolation lets you take down only that node.

What should a liquid cooling maintenance log include?

At minimum: the date and technician, every reading (temperatures, flow, pressure drop, pump speed, filter ΔP), coolant test results, makeup volume, parts replaced, and any issues found and how they were resolved. Every value should be comparable against the baseline and the previous entry.

Do sealed closed-loop modules or AIO coolers need maintenance?

Sealed closed-loop modules are filled and sealed at the factory, so users generally don't change the fluid. Keep the radiator clean and watch for changes in pump noise and temperatures. To learn how these modules work and where they fit, see what is a closed-loop liquid cooling module; the AP-AIO-240 is one APALTEK example. A field-serviced TCS loop in a data center should follow the checklist in this article.

Build Serviceability into the First Design

A maintenance checklist helps you catch problems early, but much of the maintenance workload is decided at selection: how easy the filters are to reach, where the isolation valves sit, whether quick disconnects come pre-installed, and whether the CDU has redundant pumps.

If you're planning or expanding liquid-cooled racks, send us your rack layout, coolant spec and current maintenance pain points. APALTEK's engineering team will review your CDU and cold plate loop for serviceability and recommend a solution built around it, from the CDU down to custom cold plates. We reply to quote requests within one business day.

Request a Quote