Netdata

What Netdata does

Netdata is an infrastructure monitoring agent that you install on each machine. It detects what is running, starts collecting metrics every second from the operating system, disks, network, hardware sensors, containers, virtual machines and hundreds of packaged applications such as nginx, PostgreSQL and Redis, and builds dashboards for them without any query writing. The project lists more than 800 integrations, including SNMP devices, cloud provider services, and custom applications that expose OpenMetrics or StatsD.

Two things set it apart from a typical Prometheus setup. Resolution is per second by default, so short spikes show up rather than being averaged away. And each agent trains machine-learning models for every metric locally and flags unusual behavior, which helps when you do not know in advance which metric matters. Hundreds of alert rules are preconfigured, and notifications can go to email (with a local mail server), Slack, Telegram, PagerDuty, Discord, Microsoft Teams and others. It also reads systemd journal logs on Linux and the Windows Event Log, and can run synthetic checks against APIs, TCP ports and certificates.

Licensing and the free limits

This is the part to read carefully. The Agent, which does the collection, storage, machine learning, alerting and APIs, is GPL-3.0 or later. The dashboard you see in the browser is not: the Netdata UI is a closed-source binary under the Netdata Cloud UI License, which allows use only with Netdata Agents and Netdata Cloud and forbids modification. It ships in the standard packages, so a normal install includes proprietary code.

Netdata Cloud, also closed source, is optional and adds remote access, multi-node dashboards, central alert configuration, role-based access and SSO. Without signing in, sensitive views such as processes, logs and network connections are blocked on the dashboard. Anonymous and free-tier users can view at most five active nodes at a time, and that limit applies to the local dashboard of an Agent or Parent too; the other nodes keep collecting data, but you need a paid plan to see more than five together. Dynamic configuration of collectors and alerts from the UI is also a paid feature.

Self-hosting Netdata

The recommended install on Linux is the kickstart.sh script, which uses native packages where available and a static build otherwise, and turns on daily automatic updates by default. There are also Docker images, a Helm chart, and installers for macOS, FreeBSD and Windows. No external database is needed: each agent stores its metrics in tiered storage (per second, per minute and per hour), kept for as long as disk space allows. The project gives typical use as about 5% of one CPU and 150 MiB of RAM per node with machine learning on. For central storage and longer retention, agents stream to Netdata Parents. Anonymous usage telemetry is on by default; you can disable it with an installer flag or an opt-out file.

Hosted version

Netdata Cloud has a free Community tier and paid plans. According to the project, metrics stay on your agents and parents rather than being stored centrally in the Cloud. A paid on-premises edition of Netdata Cloud is also offered.

Limitations

Netdata covers infrastructure and service metrics well, but it has no distributed tracing or application performance monitoring. The proprietary UI, the five-node viewing limit and the paid configuration features mean that a free, fully open source deployment at scale is not really possible; the GPL agent can still export data to Prometheus or other backends and be viewed in Grafana instead.

Who it suits

Netdata suits administrators of a handful of servers, home labs and teams that want detailed real-time metrics in minutes. Larger fleets should weigh the paid Cloud plans against fully open alternatives such as Zabbix or Prometheus.