Member
Overkill is relative. I run Grafana for five nodes at home. The problem is not Grafana, it is the time to make dashboard.
If you want simple, try Netdata. One command install, web UI already there, no config file to write. I use it on my mother's router in Sofia. She does not know, but I know when disk is full.
Netdata is good for "what is happening right now" but bad for "what happened Tuesday". The history is short unless you configure backend, and then you are back to config files.
Netdata is the path of least resistance, but MarchAsadi has the right warning. I ran it for a year, then hit the wall when a customer asked "show me the trend from last quarter." Nothing to show.
So I migrated to Prometheus anyway — https://prometheus.io. The sunk cost fallacy is real with monitoring.
Member
I break things so I know how they break, and I broke Netdata's SQLite backend twice. Corruption on unclean shutdown. Not production grade for me.
For ten nodes I would actually look at Datadog or New Relic, let someone else run the database. Yes, it costs. How much is your time worth?
Member
Your 'cloud' is just someone else's computer with marketing, and Datadog is just someone else's Prometheus with invoice.
For ten nodes in Germany I use Checkmk raw edition. It discovers services automatically, no copy-paste growth. Has Nagios core underneath but you do not touch it. The RAW edition is free, you pay only if you want their support or cloud.
Checkmk RAW is good. I migrated from Nagios in 2019, never looked back. The WATO config is web UI, no text files for standard changes.
But: the agent must be installed on each node. Not hard, but another step.
Member OP
Checkmk. I did not think of this. I used it once at old job, the web config was... slow, but it worked. And the discovery, yes, I remember now. You add host, it finds checks.
But I am in Lyon, the nodes are in two small datacenters near me, not AWS. Does Checkmk agent work fine on bare metal? I think yes, but I ask.
Franck, I can confirm: the Checkmk agent works fine on bare metal. I've got a Dell R620 in Lyon, no issues. Only thing: you need to open port 6556, local firewall has to allow it.
But for temperatures, Checkmk also uses the agent, not IPMI directly. If you want the actual BMC sensors, you need the separate IPMI plugin.
Member
---
- Checkmk RAW:
- cost: 0 EUR
- agent: required on each node
- bare metal: supported
- IPMI: plugin available, not default
- config method: web UI (WATO)
- Nagios heritage: yes, hidden
- Netdata:
- cost: 0 EUR
- install: one command
- history: limited without backend config
- backend: SQLite (reports of corruption)
- Prometheus path:
- cost: 0 EUR, time investment high
- history: excellent
- small scale: viable per bellaauc
- @blogfranck:
- decision tree:
- want minimal work now, accept migration later: Netdata
- want one migration forever, accept setup work: Checkmk RAW
- want modern standard, accept learning: Prometheus
---
Member
I am in Mexico City, the electricity is expensive and my servers are old. I use LibreNMS for ten nodes. It discovers via SNMP, no agent to install. The hardware health is there for Dell and HP, temperature, fan, power supply.
The interface is... from 2010, yes. But it works. And it polls, not push, so firewall is simpler.
