Field Notes
CDU alarm reset and leak response witness record before IT load increase
Before IT load is increased, a liquid-cooling record should show CDU alarm history, reset permission, leak sensor status, coolant pressure and flow trends, dew point margin, witness, photos, response owner, exceptions, and load-increase hold points.
Direct answer
Before IT load is increased after a CDU alarm or leak response event, the record should identify the CDU, rack or loop served, alarm type, alarm timestamp, cause review, reset authority, reset time, witness, leak sensor status, affected zone, coolant pressure, flow, temperature, level, conductivity or chemistry where monitored, dew point margin, pump status, valve status, dry-after photos, response owner, retest window, exceptions, and load-increase decision.
The record should prove that the alarm was not cleared as a housekeeping step. It should show who was allowed to reset it, what evidence was reviewed, what leak response actions were witnessed, and what load increase was released or held.
Use this as documentation guidance only. The CDU manual, liquid-cooling design, controls sequence, IT load plan, leak response plan, electrical safety plan, coolant chemistry plan, manufacturer instructions, owner standard, and site authority control actual reset, isolation, repair, retest, and load increase.
Why reset evidence matters
A CDU alarm can be cleared quickly in a BAS, CDU controller, or operations screen, but the risk may remain. A leak sensor may dry, a pressure alarm may recover, or a pump alarm may return to normal while the underlying cause is still unknown.
The weak record says alarm reset and no leak found. The strong record shows the alarm history, reset permission, sensor zone, inspection photos, pressure and flow trend, coolant status, witness, retest period, and exact load-increase boundary.
Trane, Vertiv, ASHRAE, TI, nVent, and liquid-cooling deployment sources all reinforce that CDU operation, leak detection, pressure, flow, temperature, and response planning are part of liquid-cooling reliability. The field record should preserve the decision before more IT load raises consequence.
Start with the load-increase boundary
The first page should name the approved load step, racks or nodes being increased, current load, target load, CDU tag, secondary loop, affected leak zone, alarm point, witness team, and hold authority.
Do not close the record at the CDU alone. A CDU may serve racks, manifolds, cold plates, rear-door heat exchangers, hoses, quick connections, and leak sensors. The record should say which downstream equipment the load increase will expose.
If the alarm occurred during commissioning, burn-in, load-bank testing, or production operation, state that context because the release risk is different.
Record reset permission
Record who authorized the reset, who performed it, what screen or controller was used, what alarm was cleared, what interlock or notification remained active, and whether the reset was temporary, permanent, or only for retest.
The permission record should include the cause status: known and corrected, known and monitored, unknown and held, nuisance alarm under review, or manufacturer support required.
Do not let reset permission become load permission. The record should separately state whether IT load can increase, stay flat, or must be rolled back.
Witness leak response evidence
If a leak alarm, wet sensor, drip, pressure loss, or coolant trace triggered the response, photograph the sensor, cable zone, drip path, under-rack area, CDU pan, manifold, fittings, hoses, quick connections, and final dry-after condition.
Vertiv leak detection and nVent leak detection sources support explicit detection, location, communication, and intervention planning. TI monitoring guidance supports leak, flow, and pressure monitoring around server cards and CDUs.
The response witness should record what was inspected, who saw it, whether any component was isolated, whether absorbent or containment was used, and what remains under monitoring.
Trend pressure, flow, and temperature
Trend CDU supply and return temperature, loop pressure, differential pressure where available, flow, pump command, pump status, reservoir or level alarm, coolant temperature, heat rejection status, and load step timestamp.
A reset without a trend can miss a recurring pressure dip, pump transition, air ingestion, flow instability, or temperature excursion. The record should show what the loop did before, during, and after the reset.
Chilldyne, Trane, Rittal, ASHRAE, and liquid-cooling deployment sources all show that liquid-cooling systems depend on controlled flow, pressure, temperature, and heat rejection. The closeout packet should treat those as load-step evidence.
Check dew point and condensation risk
If the CDU or loop controls coolant temperature near room dew point, record room dew point, coolant supply temperature, water temperature setpoint, and margin at the load-increase decision.
Lenovo water specifications and ASHRAE liquid-cooling guidance support the principle that liquid cooling near IT equipment must consider dew point and exposed surfaces. A leak alarm reset should not distract from condensation risk if colder coolant or higher load is being introduced.
Hold the load increase if the project dew point margin is missing or outside the accepted limit.
Record alarm-state normal after retest
After reset and response, record alarm status for CDU, leak sensors, BMS or DCIM points, rack sensors, pump alarms, flow alarms, temperature alarms, level alarms, conductivity or coolant-quality alarms where present, and any manually suppressed alarms.
A normal screen immediately after reset is weaker than a normal screen after a defined retest window. State the retest duration and whether the system was under no load, partial load, load bank, or IT production load.
If any alarm is suppressed, bypassed, acknowledged but active, or waiting on parts, do not present the load increase as fully released.
Record table
Use a compact table so facilities, IT, commissioning, controls, and safety teams review the same release evidence.
| Record field | What to capture | Why it matters |
|---|---|---|
| CDU and load boundary | CDU tag, racks served, current load, target load, loop, test stage | Defines what is being released |
| Alarm history | Alarm type, point, time, severity, recurrence, active or cleared state | Prevents reset from hiding the event |
| Reset permission | Authorizer, operator, reset method, cause status, retest limit | Shows who allowed the reset and for what purpose |
| Leak response | Sensor zone, photos, dry-after proof, isolation, witness, response owner | Documents whether moisture risk was addressed |
| Loop trend | Pressure, flow, temperature, pump status, level, coolant quality where monitored | Shows whether the CDU stabilized after reset |
| Condensation margin | Room dew point, coolant setpoint, supply temperature, margin | Prevents load increase into condensation risk |
| Exceptions | Unknown cause, recurring alarm, suppressed point, wet sensor, missing trend | Makes hold points explicit |
| Load decision | Increase, increase with monitoring, hold, rollback, manufacturer support | Defines the IT load boundary |
Before-load-increase checklist
Run this checklist before IT load is increased after a CDU alarm or leak response.
- CDU tag, served racks, loop, and planned load step are identified.
- Alarm history, timestamp, severity, and active or cleared state are recorded.
- Reset permission, reset operator, and witness are named.
- Cause status is recorded as corrected, monitored, unknown, or held.
- Leak sensor zone, inspection photos, and dry-after photos are included where relevant.
- Pressure, flow, temperature, pump status, and level trends are included.
- Coolant chemistry or conductivity status is recorded where monitored.
- Dew point margin is recorded when coolant temperature can create condensation risk.
- Suppressed alarms, missing points, and unresolved warnings are exception-listed.
- Load increase, monitoring period, rollback point, and owner are recorded.
Weak versus strong record
Weak record: CDU alarm reset. No leak visible. OK to increase load.
Strong record: CDU-2 low-flow alarm at 9:18 a.m. was reset by the facilities lead after the commissioning witness confirmed no active leak alarm, dry CDU pan, dry manifold fittings, normal leak cable zone LDC-2, and stable pump status. Flow and pressure trended normal for 45 minutes at the current IT load and again during a 20 percent load-bank step. Room dew point was 58.2 degrees F and CDU supply was 68.0 degrees F. Load increase was released only to Rack Row C, with a two-hour trend watch and rollback owner named.
The strong record shows reset permission, witness evidence, trend stability, condensation margin, load boundary, and monitoring owner.
Common mistakes
The most common mistake is clearing the alarm before capturing it. Once the screen returns to normal, the team may lose the point name, timestamp, severity, and recurrence context.
Another mistake is treating dry equipment as proof that the alarm was harmless. The record still needs sensor status, pressure and flow trend, witness, and retest evidence.
Other mistakes include no reset authority, no load boundary, no dew point check, no coolant trend, no dry-after photos, no suppressed alarm list, no rollback owner, and no monitoring period after load increase.
When to hold IT load increase
Hold the load increase if the alarm cause is unknown, leak sensor status is unclear, any area remains wet, pressure or flow is unstable, pump status is abnormal, coolant level or quality is in alarm, dew point margin is missing, or the reset was made without authorization.
Also hold if alarms are suppressed, the CDU manual requires manufacturer support, a fitting or hose was disturbed without retest, the response witness was absent, or the trend does not cover the load step being released.
A hold should name the CDU, rack or row, alarm point, required correction, retest evidence, response owner, and whether a lower load state can continue.
Owner handoff and monitoring
The owner handoff should include alarm export, reset authorization, photos, leak response notes, trend plots, point names, coolant status, dew point snapshot, load step, monitoring window, rollback contact, and accepted exceptions.
If the facility will continue to monitor pressure, flow, temperature, leak sensors, conductivity, pump status, or coolant level, preserve the point names and normal values from the release period.
Keep the record with commissioning and IT load-step files, not only in a service ticket.
Questions before load increase
What alarm was reset? Who had authority to reset it? What evidence says the cause was corrected or safe to monitor? What racks will receive more load? What trend proves the current load is stable?
Which leak sensors protect the CDU and served racks? Who receives the alarm? What is the rollback trigger? What coolant pressure, flow, temperature, and dew point values are normal for this release?
Answer those questions before increasing IT load.
Compliance and safety limits
This article does not authorize CDU alarm reset, leak isolation, coolant work, electrical work, or IT load increase. It is a record structure for preserving reset, leak response, trend, and witness evidence.
The CDU manual, liquid-cooling design, controls sequence, IT load plan, leak response plan, electrical safety plan, coolant chemistry plan, manufacturer instructions, owner standard, and site authority control the work. If those documents conflict with this checklist, use the controlling project document and record the decision.
Do not reset alarms, isolate valves, open equipment, handle coolant, or increase IT load outside the qualified team's authority.
Sources checked
- Trane Technologies, Coolant Distribution Unit IOMUsed for CDU safety, operation, alarms, maintenance, and trend context.
- Vertiv, ACS Cold Plate Leak Detection and InterventionUsed for liquid-cooling leak detection and response planning context.
- ASHRAE, Emergence and Expansion of Liquid Cooling in Mainstream Data CentersUsed for CDU, liquid cooling, piping, and data center integration context.
- Texas Instruments, Leak Detection, Flow, and Pressure MonitoringUsed for CDU leak, flow, pressure, and sensor monitoring context.
- Chilldyne, FAQ Guide to Data Center Liquid CoolingUsed for CDU, leak behavior, and liquid-cooling response context.
- Data Center Dynamics, Direct Liquid Cooling System ChallengesUsed for direct liquid cooling alarms, monitoring, and manual intervention context.
- nVent Raychem, Preventing Downtime in Data Centers with Leak DetectionUsed for leak detection, location, communication, and downtime prevention context.
- Dober, Data Center Leak Detection and Glycol MonitoringUsed for current leak detection, pressure, pH, chemistry, and monitoring context.
- Avtron, Liquid-Cooled Load Banks ReadinessUsed for liquid-cooled load-bank commissioning and load-step validation context.
- Lenovo, Water specifications for secondary cooling loopUsed for water temperature and dew point margin context.
- ASHRAE TC 9.9, Water-Cooled ServersUsed for facility liquid-cooling integration and monitoring context.
- Rittal, Liquid Cooling Package Plus GuideUsed for liquid-cooling package safety and alarm context.
- Vertiv, Deploying Liquid Cooling in the Data CenterUsed for CDU deployment and liquid-cooling operational context.