US Liquid Cooling editorial · Reviewed · 4 min read
Leak detection tells operators that a condition needs attention. Containment limits where liquid can go. Isolation controls the affected circuit. A reliable liquid-cooling design needs an agreed relationship between all three, rather than a sensor cable treated as the entire protection strategy.
The first design question is simple: what should the facility do when a leak is detected, and what cooling remains available while it does that?
Start with a drawing of the possible leak paths
Mark CDU connections, row and rack manifolds, flexible hoses, quick disconnects, service points and the equipment beneath them. Include the facility loop as well as the technology loop where relevant.
The consequences of a small leak above energized equipment are different from those of a leak within a contained service area. Sensor placement, drainage and isolation zones should follow those consequences, not just the easiest route for installing a detection cable.
Use product documentation for the approved detection arrangement. Vertiv's CDU 600/1350 guide specification, for example, describes leak sensing along relevant external hose or technology-loop runs.
What does a liquid sensor actually prove?
A wet sensor indicates liquid has reached that sensing location under its configured detection conditions. It does not necessarily identify the originating connection or quantify the leak rate.
Other information can help interpret the event: reservoir level, makeup activity, branch flow, pressure and temperature. An unexplained trend is not interchangeable with a confirmed wet sensor, but it may justify investigation before liquid reaches a sensing cable.
Check compatibility with the actual fluid and installation environment. Also consider nuisance events, cable damage and sensor faults. A dashboard showing “normal” is only useful when the sensing system itself is healthy.
Positive and negative pressure address the problem differently
In a circuit above ambient pressure, an opening can release liquid outward. A below-ambient-pressure cooling architecture can instead draw air inward at an opening within the protected portion of the system.
Chilldyne's negative-pressure FAQ describes that principle and its implementation. Treat its performance claims as claims about the supplier's designed system, not evidence that every negative-pressure installation has no failure modes.
Ask which parts remain below ambient pressure, including during startup, shutdown, isolation and service. Clarify how air entry is detected and handled, which faults interrupt cooling, and what components are outside the claimed protection boundary. Do not evaluate a whole facility from a single demonstration of a cut hose.
Write a cause-and-effect sequence
Agree the response before commissioning:
| Event | Response that needs an explicit decision |
|---|---|
| Confirmed liquid in a rack zone | Alarm route, affected branch isolation and IT workload response |
| Leak at the CDU | Local containment, available alternate cooling and escalation |
| Sensor disconnected or faulty | Fault alarm and operating restrictions |
| Pressure or flow anomaly without a wet sensor | Investigation criteria and responsibility |
| Maintenance on a protected section | Temporary protection and authorization to return to service |
Closing every valve immediately can protect one objective while removing cooling from unaffected IT. Conversely, continuing operation without a defined response can allow a manageable event to spread. Engineering and operations teams need to agree the sequence together.
Redundancy has to survive the actual event
Two pumps do not help if a shared leak requires the whole circuit to stop. Similarly, redundant detection electronics do not protect a hose outside their sensing coverage.
Use the specified fault case to trace what remains available: power, controls, heat rejection, fluid path and IT cooling. A platform such as Vertiv's integrated SmartIT configuration demonstrates coordinated leak monitoring in a defined assembly; the project's operating sequence still needs agreement.
What should an acceptance test demonstrate?
Use the manufacturer's safe test method to verify detection, alarm delivery, correct zone identification and the agreed response. Record timings against the project's criteria. Confirm restoration and reset behavior as well as the initial alarm.
Test a sensor fault separately from a real detection event. Verify that the alarm reaches someone who can act outside normal working hours. Keep the test evidence, isolation drawing and response procedure together.
For related planning, read liquid-cooling resilience beyond spare pumps and the commissioning handover guide. Detection is useful when it leads to a controlled, rehearsed response.
Sources & further reading
Explore the company directory → · All liquid-cooling guides →