**Netdata** is an open source real-time system performance monitoring tool that provides per-second metrics for every system resource — CPU, memory, disk I/O, network interfaces, processes, containers — with zero configuration, immediately upon installation. --- ### First Principle: Performance problems happen in seconds. Monitoring at one-minute resolution misses them. Most monitoring systems scrape every 15–60 seconds. Netdata collects at one-second granularity — revealing short-lived CPU spikes, transient network bursts, and latency outliers that [[Prometheus]] misses. This is particularly valuable during incident investigation when you need to reconstruct exactly what happened at the moment of failure. --- ### Key Considerations - **Zero Configuration**: Netdata automatically detects running services (Nginx, PostgreSQL, [[Ceph]], [[Kubernetes]] pods, etc.) and begins collecting metrics immediately — no scrape target configuration needed. - **Per-Second Resolution**: 1-second granularity is Netdata's defining feature. Data is retained at full resolution for a configurable period (default 1 day), then downsampled for longer retention. - **Anomaly Detection**: Netdata includes ML-based anomaly detection that flags unusual patterns per metric automatically — no manual threshold configuration needed for basic alerting. - **Parent-Child Architecture**: A Netdata Parent node can aggregate data from many Child nodes — providing a centralised view while Children monitor locally with full resolution. - **vs [[Prometheus]]**: Prometheus is better for long-term storage, multi-cluster querying, and custom application metrics. Netdata is better for real-time, per-node visibility with zero setup overhead. - **[[Grafana]] Integration**: Netdata can export metrics to [[Prometheus]] (via remote write) and [[Grafana]], allowing it to feed the existing observability stack. --- ### How It Fits ``` Physical/virtual nodes (any Linux system) → Netdata agent (1-second per-node metrics, auto-detected) → Netdata Parent (centralised aggregation) → [[Prometheus]] export → [[Grafana]] dashboards ``` [[Prometheus]] | [[Grafana]] | [[Zabbix]] | [[Open Source Hyperscaler MoC]]