An interim fix with no new part solved a real heat problem

Before any active cooling, the server room sat at about 90 degrees Fahrenheit and there was no way to bring it lower. The planned answer was a dedicated mini-split, already sized and quoted, scheduled for 2027. The answer that actually happened on 20260827 was to redirect a branch of cooling capacity that already existed in the building into the room, which held it around 78 degrees and cost nothing in hardware.

That part is a good story but not the lesson. The lesson is what was done next, which was to keep treating the fix as provisional.

The numbers that justified the interim approach were projections for a fleet that is not yet fully built out, so the redirect might be adequate today and inadequate at full load without anything visibly changing. Two habits keep that from becoming an under-provisioned permanent arrangement. The first is a deliberate week of pure observation before trusting the result, rather than declaring victory when the temperature first read 78. The second is keeping the mini-split sized, priced and live as a fallback rather than quietly cancelling it once the cheap fix appeared to work.

There is also a measurement trick here worth stealing. Heat load in a server room is very close to electrical load, and Node0 already polls UPS output wattage every scrape interval for entirely different reasons. Converting that to BTU gives a continuously measured heat load using an instrument already in place, with no new sensor and no estimate. If you are solving a thermal problem on a budget, look first for an instrument you already own, and treat “it seems to be holding” as a hypothesis you keep testing rather than a conclusion you file.