RackWatt: Data Center MOP

EQUIPMENT TROUBLESHOOTING, TRIPLED The reference went from six entries to eighteen, and every one carries an ordered list of what to check rather than a description of the problem. UPS gains battery string failure, running on bypass, and capacitor and fan aging. The bypass entry leads with the thing that gets missed: the critical load is unprotected right now, even though it is still up and everything looks fine from the outside. Generators gain a start then shutdown case, kept separate from a no start because fuel, battery and cranking are already proven and the diagnosis goes somewhere completely different. Plus a transfer switch that fails to transfer, which is the one where the set runs perfectly and the load stays dark. Distribution gains breaker trips, leading with not resetting into an unknown fault, and neutral heating from triplen harmonics, where the shared neutral can carry more current than any phase and a standard transformer derates badly on that load. Air cooling gains units short cycling and fighting each other, one of the most expensive faults in a legacy hall, and hot spots, which leads with measuring at the rack inlet because the room can be cold while a server inlet is not. Liquid cooling gains leak detection and CDU pressure. Both send you to the facility Risk Management Plan and Process Safety Management documentation for pressure limits rather than a remembered number. Monitoring gains alarm storms, treating alarm fatigue as the failure rather than the plant.