Put it in the runbook for the alarm you got when the server failed, or the maintainance guide that your team runs through every month. Or make heatsink repastin

Put it in the runbook for the alarm you got when the server failed, or the maintainance guide that your team runs through every month. Or make heatsink repasting a regular documented process, or something similar. Basically, as Toyota says, build quality _into_ each part of the product rather than layering it on top by documenting discrete pieces of info that an operator has to discover, understand and act upon