Logging

Observability cost is the fastest-growing line in most engineering budgets, and it grows for reasons nobody chose: debug logging left on after an incident, one label that multiplied a metric into forty thousand time series, a trace sampler set to keep everything, and a retention default nobody has read since the contract was signed. This guide separates the three signals and prices them properly, compares the pricing models vendors actually charge on, sets out benchmark ranges worth quoting in a budget conversation, works a full before-and-after model for a forty-service estate, and lists the controls that reduce spend without quietly removing the data you need at three in the morning.