Netdata

Netdata is a free, open-source monitoring agent that installs itself, figures out what’s running on your machine, and starts charting it, per second, before you’ve touched a config file. There’s no dashboard to build and no query language to learn, which puts it at the opposite end of the setup spectrum from Prometheus and Grafana.

What Netdata actually does

Costa Tsaousis built the first version of Netdata in 2013, after cloud transactions at the company where he worked started failing silently and nothing on the market could tell him why. That origin shows in what the tool still prioritizes: not dashboards, not query languages, but immediate visibility into whatever is actually happening on a box right now. Install it and it auto-discovers what’s running, nginx, PostgreSQL, Redis, and a few hundred other things, then starts charting all of it within seconds. No jobs to define, no exporters to wire up.

Two things separate it from most of the monitoring stack: collection interval and where the intelligence lives. Netdata samples per second by default, not every 15 or 30 seconds like a typical Prometheus scrape, and it trains a small anomaly-detection model for each metric locally, on the box being monitored, rather than shipping raw data somewhere else to be analyzed. The project counts 800+ collector integrations. Most of them just work without being told to.

Installing Netdata with Docker

The official image is netdata/netdata on Docker Hub, the same one Netdata’s own Docker installation docs point to. There are really two install paths: a one-line kickstart script for bare-metal or VM hosts, and Docker for anything already containerized. Docker asks for more than most containers do, because it has to see the whole host it’s watching, not just itself.

services:
  netdata:
    image: netdata/netdata:stable
    container_name: netdata
    pid: host
    network_mode: host
    restart: unless-stopped
    cap_add:
      - SYS_PTRACE
      - SYS_ADMIN
    security_opt:
      - apparmor:unconfined
    volumes:
      - netdataconfig:/etc/netdata
      - netdatalib:/var/lib/netdata
      - netdatacache:/var/cache/netdata
      - /:/host/root:ro,rslave
      - /etc/passwd:/host/etc/passwd:ro
      - /etc/group:/host/etc/group:ro
      - /proc:/host/proc:ro
      - /sys:/host/sys:ro
      - /etc/os-release:/host/etc/os-release:ro
      - /var/run/docker.sock:/var/run/docker.sock:ro

volumes:
  netdataconfig:
  netdatalib:
  netdatacache:

1. Save the compose file above and run docker compose up -d. The three named volumes keep your configuration and metric history across container recreations.
2. Open http://your-server-ip:19999. Charts should already be populating; there’s nothing to configure for this first look.
3. Use the stable tag rather than latest if you want to avoid nightly builds. netdata/netdata without a tag pulls latest by default.
4. Optional: connect the node to Netdata Cloud from Settings inside the dashboard, for a shared multi-node view and SSO. Skip it entirely and the local dashboard still works on its own.

docker compose up -d

Netdata vs Prometheus + Grafana, Zabbix, and Glances

“Netdata vs Prometheus” isn’t quite the right framing, since Prometheus is a collector and time-series database while Netdata bundles collection, storage, ML, and a dashboard into one agent. The honest comparison is Netdata against a full Prometheus+Grafana stack, or against older enterprise tools like Zabbix.

	Netdata	Prometheus + Grafana	Zabbix	Glances
Setup effort	Near zero, auto-discovers targets	Two services to configure, scrape configs to write	Server, agents, and a database to provision	Single binary, no persistence
Collection interval	Per second by default	15-30s scrape interval, typical	Minute-level, typical	Real-time, live only
Anomaly detection	Built in, trained per metric at the edge	Not built in, needs external tooling	Trigger-based thresholds, not ML	None
History retention	Local disk, Cloud/Parents for centralizing	Configurable, commonly weeks	Long-term, database-backed	None, live snapshot only
Best for	Fast, detailed single or few-host visibility	Custom alerting, service metrics, scale	Larger, structured IT environments	A quick terminal-based check

Netdata: pros and cons

• Auto-discovers services and starts charting them immediately, no dashboards to build first
• Per-second collection catches spikes that a 30-second scrape interval would smooth over and miss
• Anomaly detection is trained per metric locally, no separate ML pipeline to stand up
• One agent replaces the collector-plus-dashboard split that Prometheus and Grafana require

• The full Docker deployment needs host-level privileges (pid:host, network:host, SYS_ADMIN) that most containers never ask for
• No login on the dashboard by default, same exposure risk as Prometheus without a proxy in front
• Retention is local-disk by default; centralizing history across several nodes really wants Netdata Cloud or a self-hosted Parent
• The three-way license split between Agent, UI, and Cloud trips people up, since “open source” only accurately describes part of it

FAQ

Is Netdata free?

The Agent is, under GPLv3+, with full source on GitHub and no feature gate. The dashboard UI is free to use but closed source. Netdata Cloud adds a free community tier plus paid tiers for larger teams that want SSO and centralized management, none of which is required to self-host.

Do I need Netdata Cloud to use it?

No. The local dashboard at port 19999 works standalone, no account or sign-in required. Netdata Cloud only matters once you want one view across multiple nodes, or SSO for a team.

How much CPU and RAM does Netdata use?

Around 5% CPU and 150MB RAM by default, per the project’s own FAQ, less than 1% CPU and about 100MB RAM with ML and alerting disabled. That’s for the Agent itself, separate from whatever it’s monitoring.

Is Netdata better than Prometheus and Grafana?

Different tradeoff, not a strict upgrade. Netdata gets you detailed, per-second visibility into one or a handful of hosts with almost no setup. Prometheus paired with Grafana takes more work up front but scales further and gives you real control over alerting logic through PromQL. Plenty of homelabs run both for different jobs.

Can Netdata monitor Docker containers directly?

Yes, built in, through a read-only mount of the Docker socket, no separate exporter needed. Prometheus needs cAdvisor bolted on for the same job; Netdata just needs the socket mount already shown in the compose file above.

Netdata is the fastest way to get real, per-second visibility into a single box, or a handful of them, without writing a scrape config first. It’s worth pairing with Uptime Kuma for the simpler up/down question Netdata isn’t built to answer, and with Prometheus and Grafana once a homelab grows past a couple of hosts and needs custom alerting rules. Our homelab setup guide for beginners covers the Docker basics it runs on, and the Monitoring category rounds out the rest of the tooling here.