Observability with Prometheus and Grafana

Reliability Over Completeness

A metaphor that can, in my opinion, explain how Prometheus approaches data collection is that of a news reporter covering a breaking story. They might not have all the details yet, but they can still give you a decent understanding of what's happening.

Prometheus works in a similar way. It prefers reliability over completeness. This means that even if it doesn't capture 100% of data points, it will prioritize providing you with a consistent view of your system's health. In most monitoring scenarios, a few missing data points are acceptable if it means you can still get a good overview of what's happening.