Top-of-rack hot spots at the far end of a contained hall

What the data rules out. Installed capacity is not the problem: five CRAH units remain available with one down (400 kW) against a 380 kW load, chilled-water ΔT is a healthy 4.8 K, and fans run at only 62%. A capacity shortfall would show elevated inlet temperatures along the entire row and a rising return-air temperature, not a localised top-of-rack effect confined to four racks.

What the data supports - recirculation plus starvation.

1. Missing containment roof panels above racks 20–21. Hot-aisle air is buoyant and accumulates under the containment roof. With the panels removed, that air spills over the top of the racks back into the cold aisle, so the top 3U of the neighbouring racks draw a mixture of supply air and return air. This explains why the worst readings are at the top U and why the effect only appears at peak load, when the hot-aisle/return temperature differential is largest.
2. Low underfloor pressure. Orifice flow through a perforated tile scales roughly with the square root of the plenum-to-room pressure difference. At 8 Pa versus a 22 Pa design, tile flow at that end of the hall is a fraction of nominal, so the far racks are under-supplied at every height - and the top U suffers most because that is where recirculated air collects.
3. Missing blanking panels at rack 22. Empty U space is a bypass path: cold air escapes through the rack straight into the hot aisle without cooling any IT, which both wastes supply air and raises hot-aisle pressure, pushing more air back over the top of the containment.

Ranking. Containment breach first (largest, most localised), plenum pressure second (it affects the whole far row), blanking panels third - but they interact, so all three must be corrected before re-measuring. Note that 31–33 °C exceeds the ASHRAE A1 recommended maximum of 27 °C and sits at the A1 allowable limit of 32 °C, so this is an IT-reliability exposure, not a comfort complaint.

Phase 1 - immediate, no outage, same day

• Refit the two containment roof panels above racks 20–21 and re-seal all cable penetrations through the containment (brush strips, grommets, end-of-row doors).
• Install the 14 missing blanking panels at rack 22 and sweep all 22 racks for further gaps.
• Verification: thermal-imaging survey of the containment roof line plus fixed-point measurements at U1, U20 and U40 of racks 19–22, taken at the daily load peak over at least 72 hours. Acceptance: no top-U reading above 27 °C with the room at design load.

Phase 2 - airflow balancing, one maintenance window

• Inspect the plenum for leaks, open cable cut-outs and missing grommets; check that tile dampers at the far end are fully open and that tiles have not been replaced with solid panels.
• Measure tile flow with a balometer at a sample of positions in each row and adjust dampers or relocate tiles until the far-end plenum pressure is at least 18–22 Pa at design flow.
• Verification: repeat the pressure survey and the rack-inlet survey; target a spread of no more than 3 K across all rack inlets.

Phase 3 - residual risk

• If any rack still exceeds 27 °C at design load after Phases 1–2, add local cooling (rear-door heat exchanger or in-row unit) at racks 19–22 rather than raising hall-wide airflow, which would be energy-inefficient.
• Update the change-control procedure so that containment panels removed for cabling are recorded on a permit and reinstated and signed off before the permit is closed - the root cause here is a process gap, not a design error.

Reject it as presented. The proposal confuses two different problems.

1. It makes the hot spot worse, not better. Cooling capacity delivered to the room is proportional to mass flow times the supply-to-return temperature difference. Raising the supply setpoint by 3 K while airflow is unchanged reduces the capacity each CRAH delivers and simultaneously raises the temperature of the air available to dilute recirculated hot air at the top of the racks. The racks already sit at 31–33 °C; a 3 K uplift on the supply side pushes them further past the ASHRAE A1 recommended maximum of 27 °C and toward the 32 °C allowable limit, where IT vendors may decline warranty support.

2. The room is not over-cooled. With mid-rack inlets at 24–25 °C, the hall is already at the design point. There is no surplus cold air to give away; the 8 Pa plenum reading says the air is not reaching the far racks in the first place.

3. It masks the fault. Raising the setpoint would flatten the apparent temperature spread on the monitoring dashboard while leaving the containment breach and the unbalanced plenum in place, making the next incident harder to diagnose.

What to do instead. Fix distribution first (roof panels, blanking panels, plenum pressure). Once the hall is uniform and every rack inlet is comfortably inside the recommended envelope, a setpoint increase can be evaluated properly - with a documented baseline, a CFD or measured airflow study, and a stepwise change of 0.5–1 K at a time, verifying rack inlets after each step. At that point the change is a legitimate energy measure and can be justified against PUE and CRAH fan/pump energy data rather than used as a workaround for an airflow defect.