Life was apparently a lot easier back in the mists of time. We monitored single hosts running single services. Each their own central and global state. Easy to probe, easy to watch, easy to collect and correlate information. Easy to isolate faults. This simplicity, however, was not reflected in the quality of our monitoring. Despite life being easier, we still lived in a dark time for monitoring. We monitored the wrong components, in the wrong places, in the wrong ways, and with the wrong customers in mind. Why the contradiction, and what can we learn from it?