Navigating Coverage-First vs. Cost-First in Adaptive Metrics

In the world of observability, efficiently managing metrics is crucial. Adaptive Metrics is designed to analyze how your metrics are utilized and provide recommendations on the level of detail to maintain for each metric. This capability helps you optimize your data collection, ensuring you have the right information without overwhelming your system with unnecessary data.

Adaptive Metrics offers two primary modes for applying its recommendations: coverage-first and cost-first. In coverage-first mode, every recommendation is applied immediately, regardless of whether it increases or decreases your series count. This approach maximizes the coverage of your queries, ensuring you capture as much relevant data as possible. On the other hand, cost-first mode introduces a safeguard by only applying recommendations that won't increase your total active series. It operates in three steps: it applies series-reducing recommendations immediately, ranks series-increasing recommendations from least to most impactful, and stops applying them once the projected series count reaches its original level. This careful consideration mimics a manual review process, allowing for a more controlled application of changes.

When implementing Adaptive Metrics, you'll need the Admin role to make changes. The General Availability release of auto-apply on 2026-10-15 marks a significant step forward in automating metric management. However, be aware that while coverage-first is straightforward, cost-first requires a deeper understanding of your metric landscape to avoid unintended consequences. The balance between data richness and system performance is delicate, and knowing when to apply each mode is key to successful observability.