Netdata in an Afternoon: Per-Second Metrics, Zero Config

Thu, 11 Jul 2024 15:55:00 +0000

Most monitoring setups sample every fifteen seconds. Prometheus defaults to it, and for a great many things fifteen seconds is fine. Then one day a service stutters — a CPU spike that lasts four seconds, a disk that saturates for a moment, a burst of network that fills a buffer — and your fifteen-second dashboard shows a serene flat line, because the event began and ended entirely between two samples. You know something happened. You have no picture of it. The resolution ate the evidence.

Netdata - vo.rs

Netdata in an Afternoon: Per-Second Metrics, Zero Config