DeviceMonitor
A small self-hosted dashboard that watches every device in my home network, from the router to the NAS boxes and the UPS, and tells me when something is wrong.
The idea
At home I have a router with two internet lines, several NAS boxes, a small server, a couple of Raspberry Pis, a PC that is almost always off, a UPS and a second network at a remote house. Each has its own web page, its own quirks, and none of them tells me anything until I go and look. DeviceMonitor is one page that shows them all, and warns me when something needs attention.
Design choices
- Standard protocols, no agents everywhere. Each device is read with whatever it already speaks: the usual network-management protocol (the secure version only), a small metrics exporter, the UPS tools, the router’s own interface. Every kind of device gets a profile, so adding a new one is mostly configuration.
- Read-only. The monitor never changes anything on the devices it watches, and each one has its own read-only account.
- A device that is off is not an incident. Some machines are switched off on purpose or on a schedule; they are marked as such, and the expected off-hours of each one are configured, so it stays quiet in that window and only complains if it is off at the wrong time.
- Alerts with a delay. Each type of problem has its own waiting time, and a device can override it (a NAS that briefly touches a high temperature under load waits fifteen minutes before it complains), so a short glitch does not wake me up. Alerts go out as a chat message and as an email, and the state is saved so a restart does not repeat them.
- Who watches the watcher? A second, independent machine checks every few minutes that the monitor itself is alive, and writes to me if not.
- Private by default. It is reachable only from the home network and from the VPN, over HTTPS with a proper certificate that renews itself.
What’s built so far
- Phase 1: collectors, a small archive of measurements and a web interface, with the first devices: the server and its UPS.
- Phase 2: the router, with the state of both internet lines, and the NAS boxes: system, disks, volumes and their UPS.
- Phase 3: alerts by email and chat, and the external check.
- Phase 4: history with charts for 24 hours, 7 days, 30 days and a year on each device page.
- Then HTTPS, the PC that is often off, the router at the remote house (with its mobile signal quality) and the two DNS servers, including the free space on their memory cards.
- Last, the VPN server (with the state of the tunnels) and the home-automation Pi (with the state of its services). Small exporters there are password-protected.
- On 15 November the two media NAS boxes: the state of their disk arrays and a check that the volumes that must be mounted really are.
- On 16 November the expected off-hours for the NAS boxes with a schedule, per-device alert delays and the health of the disks inside a hardware disk box that hides its own state. The monitor read them once a day and found an old disk with warning signs, so the plan is to copy its data elsewhere and replace it.
- Last, the two backup-only NAS boxes, on a lightweight exporter that runs without containers, and a fix for a NAS whose clock was drifting because its own time sync never ran while it was off.
- On 19 November two new checks: the expiry of the HTTPS certificates of my home services, so a failed renewal warns me days ahead, and silent timers, meaning scheduled jobs (backups, renewals) that quietly stop running. Plus a per-device option to ignore unreliable readings. I used it for one NAS’s UPS that reported a charge stuck at zero; that same evening the UPS was fixed, it read the right charge again and the ignore was removed. A useful option, but one to keep for real cases.
- About sixteen devices monitored, and a test suite that keeps the collectors honest.
Status
Running and in daily use since 14 November 2026, and now complete: every device I wanted to watch is covered, including the two backup-only NAS boxes that I first left out because they are often off on a schedule. Once their off-hours were configured and their clocks fixed, they fit in too. While building it I also found weak spots in the network setup itself and closed the services that were needlessly exposed to the internet, a nice side effect.
Stack
Python · FastAPI · SQLite · Docker