Layer 3 · self-hosting reality check
What it actually takes to self-host Netdata
The docs say 200 MB. In practice you want 1 GB per monitored node. Here is the honest version — real requirements, real monthly cost, what you will be maintaining, and the one thing that catches people out.
Usually reached from Datadog alternatives, where Netdata is one of the picks.
Wondering whether you need to at all? Is Datadog free? — what the free tier actually allows, and where the wall is.
| RAM — documented minimum | 200 MB |
|---|---|
| RAM — what it really needs | 1 GB per monitored node |
| CPU | 1 vCPU — the agent is genuinely light |
| Disk | Tiered storage; a few GB per node per month at default retention |
| Monthly cost | $6/mo for a parent node collecting from agents, against per-host APM pricing that starts around $15 per host per month |
| Setup time | 15 minutes — one install script and it is already collecting |
| How you install it | one-line kickstart script, or a container per node |
| Ongoing maintenance | Low. The defaults are unusually good; the work is deciding retention and alert routing. |
| Where it stops scaling | Hundreds of nodes streaming into a parent. Per-second granularity across a whole estate is the thing nothing else gives you at this price. |
The thing that catches people out
Out of the box every node runs standalone with its own dashboard and its own alerts, so a twenty-node estate sends you twenty separate alert streams and nobody reads any of them. Configure parent-child streaming on day one so metrics land in one place — retrofitting it after alert fatigue has set in is how people abandon Netdata and conclude it is noisy.
When not to self-host Netdata
You need long-retention historical analysis. Netdata's strength is per-second resolution over a short window; it is a real-time troubleshooting tool, not a warehouse for capacity planning.
Every guide here carries this section. A site that only ever tells you to self-host is selling something — the useful answer is sometimes no.
Other Layer 3 self-hosting guides
Common questions
- How much RAM does Netdata actually need?
- 1 GB per monitored node in practice. The documented minimum is 200 MB, which is the figure at which the process starts rather than the figure at which it works under real use. 1 vCPU — the agent is genuinely light alongside it.
- What does self-hosting Netdata cost per month?
- $6/mo for a parent node collecting from agents, against per-host APM pricing that starts around $15 per host per month This is commodity VPS pricing and excludes your time, which is the larger cost for most people — budget for low. The defaults are unusually good; the work is deciding retention and alert routing.
- How long does it take to set up Netdata?
- 15 minutes — one install script and it is already collecting, via one-line kickstart script, or a container per node.
- When should I NOT self-host Netdata?
- You need long-retention historical analysis. Netdata's strength is per-second resolution over a short window; it is a real-time troubleshooting tool, not a warehouse for capacity planning.
- What is the most common mistake when self-hosting Netdata?
- Out of the box every node runs standalone with its own dashboard and its own alerts, so a twenty-node estate sends you twenty separate alert streams and nobody reads any of them. Configure parent-child streaming on day one so metrics land in one place — retrofitting it after alert fatigue has set in is how people abandon Netdata and conclude it is noisy.