Hot aisle and cold aisle containment inspection checklist: every item, the reading that proves the seal held, and the release test the fire code adds

Row 14's cold aisle was contained in March. In August a rack at position 14-07 was swapped for a shorter chassis, the cable pull through the floor cutout was never re-sealed, and at 02:00 the inlet at the top of 14-06 reads 29 °C while the CRAH supply has not moved. The weekly containment walk was done on Tuesday and every box on it is ticked.

A hot aisle and cold aisle containment inspection checks that the barrier is closed and intact: end-of-row doors latched, roof and vertical panels seated, blanking panels in every empty rack unit, cable cutouts sealed, and perforated tiles only in the cold aisle in front of loaded racks. It then checks the result against ASHRAE's recommended 18 to 27 °C inlet band and confirms the fire detection and suppression still reach the contained aisle. Each item on this list also carries the reading that proves it held and the event that reopens it, because a containment leak is a change somebody made, not a part that wore out.

• The barrier is mostly rack. Trade press puts racks at roughly 60 percent of the containment surface, so blanking panels and cable seals are the list, and the doors and roof are the rest.
• Before containment, Uptime's measurements found 60 percent of supplied cold air returning without passing through IT equipment, 61 percent of it through unsealed cable cutouts and 39 percent through tiles outside the cold aisle. Containment is the answer to that number, and every unsealed change reopens it.
• Three readings say whether the seal held: the rack-inlet spread against 18 to 27 °C, the Rack Cooling Index at or near 100, and the Return Temperature Index near 100, below it for bypass and above it for recirculation.
• NFPA 75's 2024 edition lets a site keep its sprinkler and clean-agent layout unmodified only if smoke detection removes the containment obstruction before suppression operates, on a listed releasing device, for the whole zone. That release is a test with a date, and no page-one checklist carries it.

What is a hot aisle and cold aisle containment inspection?

Aisle containment is the set of physical barriers that stop supply air and exhaust air from mixing in a hot aisle and cold aisle layout. Energy Star describes the rigid form as turning either the cold aisle or the hot aisle into its own room by sealing it with doors, sidewalls and roofs, with strip curtains as the inexpensive version and chimneys as the ducted one. Cold aisle containment makes the room the hot side, and hot aisle containment makes the room the cold side and ducts the exhaust to the coolers or a ceiling plenum.

The inspection is a walk of that barrier plus a read of the sensors behind it. The walk finds the propped door and the missing panel, and the sensors find the leak nobody can see, which is why a list that stops at the tick is only half an inspection.

End-of-row doors

Swinging or sliding, single or double, with a closer, a latch and emergency break-out hardware. The largest single opening in the aisle when propped.

Roof and vertical panels

The aisle ceiling, the rack-top to ceiling infill and the row-end fillers. Removable panels sit in releases that may be tied to the fire alarm.

Strip curtains

The flexible version of both, hung on a track. Strips go missing one at a time.

Blanking panels

Fill every empty rack unit so exhaust cannot re-enter the rack and reach the inlets above.

Grommets and brush panels

Grommets seal the vertical path through a floor cutout, brush panels seal the horizontal cable entry at the rack face.

Perforated tiles

The supply openings. One in front of each loaded rack, solid tiles in front of empty racks and everywhere outside the cold aisle.

Aisle containment is also a fire-code term. NFPA 75's 2024 edition defines it as an HVAC method in the occupied area of an air-cooled ITE space that physically separates hot exhaust air from cooler intake air, and gives it its own section, 6.7, on materials, detection and suppression. The inspection below carries both readings of the word, the thermal one and the fire one, because the same panel serves both.

Why does a contained aisle need inspecting after it is built?

Because the reason it was built is a number that comes back. The Uptime Institute study that named the problem measured 19 raised-floor rooms and found that 60 percent of the cold air supplied was short-cycling back to the cooling units without passing through IT equipment, with 61 percent of that waste going through unsealed cable cutouts and 39 percent through perforated tiles placed outside the cold aisle. Each unsealed cable opening short-cycled the airflow of half a perforated tile, and the highest share of hot spots was in the most lightly loaded rooms.

Containment is the structural answer to that. Schneider's white paper 135 puts typical raised-floor leakage at 25 to 50 percent against 3 to 10 percent for a containment system, and that gap is the whole return on the installation. The gap is made of doors, panels, blanking plates and grommets, which are the things people move when they move a rack.

A containment leak is not wear. It is the last rack swap, cable pull or tile lift that nobody sealed behind.

Uptime's own journal makes the second point. Its 2014 review found an average of only 40 percent of the cooling air supplied being used to cool IT equipment, cooling capacity ratios that had worsened from 2.6 to 3.9 times the IT requirement, and 57 percent of surveyed sites controlling on room or return air, which it calls the two least effective control points. A contained aisle whose CRAHs still control on return air has a seal and no way to bank it.

Sealing is also not the end of the job. Lars Strong's 2025 note for Upsite argues that grommets, tiles and blanking panels are prerequisites that alone do not reduce bypass airflow, because the sealed air comes out of the perforated tiles instead, and only a change to the volume the cooling units supply changes the bypass. That is why the checklist below ends in readings rather than ticks: the readings are what say the seal was banked.

What is on the containment inspection checklist, and what proves each item?

The table is the list, assembled from Energy Star's guidance, Uptime's best-practice set, the component vocabulary of the containment vendors and NFPA 75's containment section. Two columns are added to what any of those carry: the event that reopens the item, and the reading that proves it held.

Two things follow from the third column. The list has no natural cadence of its own, because nine of the fourteen items are reopened by a rack move, add or change, so the honest interval is every change plus a walk often enough to catch the ones nobody logged. And the unit-side items are not here on purpose: filters, fans and coils belong to the CRAC and CRAH maintenance list, and mixing the two produces a walk that ticks the unit and misses the aisle.

Which readings show the containment is leaking before anyone sees the gap?

The walk finds what a person can see. The readings find the leak, and they find it in the right place, because LBNL's measurement guidance is blunt that most ambient temperatures have very little to do with the actual IT equipment intake air temperatures. The same guidance places sensors at the intake of the top, middle and bottom equipment in a rack, flush with the perforated door so they move with it, on every second or third rack along the row.

Bypass

Conditioned air that returns to a cooling unit without passing through IT equipment. It leaves through cutouts, gaps under racks and tiles outside the cold aisle. Return Temperature Index below 100.

Recirculation

Exhaust that re-enters an inlet. It comes through empty rack units, door gaps and the space over a short rack. Return Temperature Index above 100, and the top-of-rack inlet hottest.

The Return Temperature Index is the reading that does most of the work, and it is older than any containment vendor. LBNL's benchmarking guide defines it as ideal at 100 percent, when the return air is at the temperature leaving the IT equipment and the supply is at the rack-inlet temperature, and reads a value under 100 as supply bypassing the racks and a value over 100 as recirculation from the hot aisle. A sealed aisle that still runs well under 100 has a seal and an unbanked one.

What does NFPA 75 require of aisle containment, and what is the release test?

NFPA 75's 2024 edition, issued by the Standards Council on 1 December 2023, carries the containment requirements in section 6.7 and the detection and suppression requirements in chapter 9. The edition is the one in force, and the clause numbers below are its own, not the 2013 numbering that still circulates in vendor articles.

The exemption is the clause that matters for an aisle added after the room was built, which is most of them. It lets the sprinkler and clean-agent layouts stand as they are only when every one of five conditions holds, and the first condition is the one a heat-released drop-away panel cannot meet.

1

An automatic means of smoke detection initiates the removal of the obstruction before the suppression system operates.

2

Removing the obstruction, or part of it, does not compromise the means of egress per NFPA 101.

3

The design and installation of the removable elements does not diminish the level of protection that existed before the containment was installed.

4

The releasing devices are listed for the application.

5

All removable obstructions are removed for the entire suppression zone.

The annex is explicit about the alternatives. A.6.7.8.1 says that means other than automatic smoke detection for removing the obstruction, and it names thermal, mechanical and fusible links, still need further research and are not clearly demonstrating the capability of activating without impacting the timed response of suppression systems. A fusible link has to melt, and a panel that waits for that is a panel that opens after the sprinkler wanted it open.

NFPA 75 does not set a test interval for the release itself. It says the systems are tested as necessary and hands the releasing device to NFPA 72 and its listing, which puts the test on the site's fire-alarm inspection cycle and in the fire-alarm contractor's hands. What the containment checklist owns is the row: the date of the last release test, who signed it, and whether every removable panel in the zone dropped.

How to seal a finding without overheating a rack

A seal is a change to the room's airflow, and the study that started all of this is careful about the order. Its authors write that in an operating room remediation should be considered only after completing a baseline study of how cooling is actually occurring, because closing too many openings in the wrong order or too quickly can upset the ambient cooling that was quietly carrying the equipment and overheat it. Their own case study took bypass from 43 percent to under 10 and the hottest intake from 30 °C to 21 °C, in that order.

• Take the baseline before touching anything: every rack-inlet channel in the row, the CRAH supply and return, the supply setpoint, and the Rack Cooling Index and Return Temperature Index as they stand.
• Walk the row against the table above and log each open item with its trigger, so the record says which rack swap or cable pull opened it.
• Seal from the exhaust side inward: blanking panels and rack-face entries first, because those are recirculation and they raise the top-of-rack inlet directly.
• Then the floor: grommets in the cutouts, tiles moved back to the cold aisle and solid tiles in front of empty racks, one aisle at a time, with the inlets watched between each step.
• Then the barrier: doors, closers, roof panels, infill and curtains.
• Hold the observation window. ASHRAE allows no more than 5 °C of inlet change in any 15 minutes, and a fast seal on a starved aisle can move an inlet faster than that in the wrong direction.
• Only then bank it: with the inlets steady inside the band, a person raises the supply setpoint or trims the fan speed, and the Return Temperature Index moves toward 100.
• Close the finding when the post-seal readings match the proof column, and leave it open when they do not.

The banking step is the one the checklists never reach, and it is the one Upsite's 2025 note is about. Sealing a cutout sends the same air out of a perforated tile instead, so the room's bypass does not fall until someone changes the volume the cooling units supply, and that is a setpoint or a fan-speed decision a named person makes at the controller with the inlets in front of them.

What do the BMS, DCIM and CMMS already hold about containment?

Each system holds one column of the checklist table, and none holds the join between the change, the item it reopened and the reading before and after. The gap is not that the systems fail to communicate. It is that the change is logged in one, the reading trends in another, and the tick lives in a third that never sees either.

The join is cheap once the trigger column exists. A move, add or change in DCIM that names a rack in a contained row is the event that reopens nine of the fourteen items, and the BMS already has the inlets that will say whether anyone sealed behind it. What is missing is the record that says the change happened, the item reopened, and here is the reading that closed it.

Who may work in the hot aisle, and who may test the release?

A contained hot aisle is a workplace at a temperature the rest of the room never sees. Schneider's white paper works an example in which a 25 percent work and 75 percent rest regimen keeps the wet-bulb globe temperature under 32.2 °C and lets the hot aisle run as high as 47 °C, and the trade press quotes hot aisles at 35 to 43 °C. OSHA's guidance says workplace heat should be measured on site with a wet-bulb globe thermometer, gives NIOSH-adapted limits of 30 °C for acclimatised light work and 28 °C for unacclimatised, and says to treat anyone on the job under one to two weeks as unacclimatised.

Two of the fourteen items are not the facilities technician's to sign. The release test is a test of a listed releasing device on the fire-alarm system, and NFPA 75 hands it to NFPA 72, which puts it with the fire-alarm contractor on the site's test cycle. The detector listing in the hot aisle is the same contractor's record, and which credential the person on shift actually holds decides who walks the aisle, who moves a tile, and who signs the fire row.

The equipment sets its own rule for the walk. ASHRAE's guidelines say that during maintenance on any hardware a properly functioning and grounded wrist strap must be worn by anyone who contacts ITE, and they give a separate powered-off envelope for equipment that is installed but not in use during repair or upgrade. Fitting a blanking panel is contact with the rack, and the strap is part of the item.

What this cannot do

A checklist assembled from government guidance, Uptime's research and the fire code is not your aisle's as-built. The bands here are ASHRAE's, the indices are LBNL's, the clauses are NFPA 75's 2024 text, and none of them knows where your releases are or what your fire-alarm contractor's cycle is. The bypass figures are from raised-floor rooms measured before containment was common, and they are the reason for the seal, not a description of a sealed room.

Nothing here takes control of a door, a tile, a setpoint or a releasing device. A decision layer that reads the inlets and the change record can say that rack 14-07 was swapped on Tuesday and the top of 14-06 has been outside the band since, and that the finding should not close, and it does not open, move, seal, rebalance or test anything. A named person fits the panel, a named person raises the setpoint with the inlets in front of them, a listed device on the fire alarm drops the roof, and the layer records what was read before and after and whether the two agreed. The ride-through the aisle has when a cooler fails is a separate question, and containment can shorten it as easily as lengthen it.

Answered