We Ran Three Free Network Monitors on a $45 Raspberry Pi and Lied to Them for a Week

The pitch was simple: take a $45 Raspberry Pi 4, install Zabbix, LibreNMS, and Uptime Kuma, then let all three watch a fake small-office network for seven days. We wanted to know which one a one-person IT shop could actually live with—not which one has the longest feature list. So we built a tiny network in the lab, staged three outages, and counted every alert like it was coming out of our own pocket.

The hardware, and the setup tax

Our test box was a used 4GB Raspberry Pi 4 from Micro Center, a 32GB Samsung EVO SD card, the official power supply, and a small fan case. All in, about $70. No UPS, no cluster, no cleverness. It sat on the shelf next to a power strip and a label maker, which is exactly where a lot of small-shop monitoring ends up.

Setup time was the first surprise:

• Uptime Kuma: 12 minutes. Docker run, web UI on port 3001, SQLite, done. It handles ping, HTTP, TCP, and DNS checks out of the box. No SNMP.
• LibreNMS: 1 hour 40 minutes. Ubuntu 24.04, MariaDB, SNMP, the works. Discovery found our fake switch and drew port graphs without much arm-twisting.
• Zabbix: 2 hours 15 minutes. MySQL, server, frontend, agent, templates, triggers. Zabbix is the one that asks you to mean it. If you skip trigger tuning, it will happily tell you about every consequence of every problem.

That difference matters. A free monitor that eats your Saturday is still free, but it is not cheap.

How we faked the outages

We stood up an OpenWrt router VM, a Linux box pretending to be a 24-port switch with snmpd, a dnsmasq server, and a WAN uplink we could throttle or blackhole. Then we scheduled three events:

1. Switch failure: one access port went down on Tuesday at 10:14 a.m. for six minutes.
2. DNS failure: port 53 went into a blackhole on Thursday at 2:03 p.m. for eleven minutes.
3. WAN failure: the default route dropped on Friday at 8:47 a.m. for four minutes.

We also added a slow-burn event: twenty minutes of high latency on the WAN link, because real networks rarely fail cleanly.

Dashboard usability: the glance test

We judged each dashboard by how fast a tired person could answer one question: what is broken right now?

Uptime Kuma wins the glance test. Big colored pills, clear uptime percentages, a heartbeat bar that looks like a hospital monitor. You do not need training. You need eyes.

LibreNMS is the network closet dashboard. Port graphs, interface errors, bandwidth, device health. If you have a switch with a flapping uplink, LibreNMS shows you the port and the traffic spike without making you build a query. It is less friendly on a phone, but it is the one we would want while standing in a rack.

Zabbix is the most powerful and the least forgiving. The default dashboard is a wall of widgets, and the “problems” screen is where you will spend your time. It can do almost anything, including agent-based monitoring and dependency-aware triggers, but it feels like a tool built for people who monitor networks for a living. That is not an insult. It is a warning label.

Alert noise scoreboard

This is where the week got interesting. We routed every notification to email and a free Discord webhook, then counted.

• Uptime Kuma: 6 alerts. One per real event, plus a recovery notice. Quiet, clean, predictable.
• LibreNMS: 19 alerts. The port-down event triggered interface and service alerts, and the DNS blackhole produced repeated failures until we tuned the delay. Still manageable.
• Zabbix: 31 alerts before tuning. The switch failure alone set off dependent service triggers because we had not configured dependencies. After about forty minutes of trigger tuning, it dropped to 9. That is a configuration problem, not a Zabbix problem—but it is also a Tuesday problem.

Alert fatigue is the hidden cost of “free” monitoring. A tool that cries wolf gets muted, and a muted monitor is just a fancy clock.

Notification costs: free software, paid plumbing

All three tools are free. The notifications are where money sneaks in.

Email through a free SMTP tier or a Gmail app password costs nothing. Discord and Slack webhooks cost nothing. Pushover is a one-time $5 per platform, which is hard to beat. SMS is the trap. Twilio charges roughly $0.0079 per outbound US SMS before carrier fees. If your monitor sends twenty texts a day, you are looking at a few bucks a month—not huge, but not zero. And if you wire up a pager service, the monthly seat fee can eclipse the cost of the Raspberry Pi in a quarter.

Uptime Kuma supports a ridiculous number of notification services, including SMS integrations. LibreNMS and Zabbix can do it too, but you will likely bolt on a mail relay or webhook. The monitor is free; your alert path is a small budget line.

Resource use and SD card reality

Running all three at once is not something we would recommend. Average CPU on the Pi was about 2% for Uptime Kuma, 11% for LibreNMS, and 18% for Zabbix. RAM landed around 180MB, 420MB, and 610MB respectively. Together they fit in 4GB, but polling spikes made the Pi feel busier than the numbers suggest.

We also moved the databases to a cheap USB SSD after day three. SD cards and databases are not friends. If you run any of these long-term, spend the $9 on an adapter and a spare SSD. Your future self will thank you.

The verdict

For a small shop with five to fifty devices and no dedicated monitoring person, Uptime Kuma is the one we would leave running. It is not a network management platform, and it does not pretend to be. It tells you when something is down, and it shuts up when things are fine.

If you need port-level graphs and SNMP discovery, LibreNMS is the sweet spot. It took longer to set up, but the payoff is real visibility into switches and links without a license fee.

Zabbix is the right answer for agent-based monitoring, complex triggers, and environments where someone actually owns the monitoring stack. Budget the setup time and the tuning time. Do not install it on a Friday and expect a quiet weekend.

None of these tools cost a subscription. All of them cost attention. The Pi is still running Uptime Kuma on the shelf. The other two are SD card images in a drawer, which is probably where they belong until the network gets bigger.