Get an Alert When a Device Goes Offline on Your Network

The camera stopped recording three days ago. The NAS backup didn't run because the NAS was off. The freezer sensor lost Wi-Fi. You find out when it's too late, because nothing told you. In this video we set up an alert for exactly that: a message on your phone when a device on your network goes offline, and only when it matters.

We use DeviceShelf Server, which runs on an always-on machine at home and checks the network every few minutes. What you see is its demo network, a made-up home with forty six devices, so none of the addresses are real.

Open Settings, Alerts and notifications. At the top is the list of events the server can report. The one we want is: a device goes offline. Next to it: a device comes back online, so you also know when it's fixed.

The next question is the important one: when does a device count as offline? Instant means after one missed check. That's quick, but Wi-Fi devices miss a check now and then without anything being wrong. Standard waits for three missed checks in a row, which filters out most false alarms. Start with that.

On top of that you can say how long a device must stay offline before the alert goes out, in minutes. Five or ten minutes is a good value for things that restart now and then, like a router or a NAS doing an update.

Phones and tablets are a special case. They drop off the Wi-Fi every time someone leaves the house or the screen locks. This switch keeps their coming and going out of your alerts, unless you want a log of who is home.

Now, where should it go? Under Notifications, add a channel. The list has push apps like ntfy, Gotify and Pushover, chat apps like Discord and Telegram, and many more. E-mail has its own section below.

ntfy is the quickest start. Install the free ntfy app on your phone, subscribe to a topic name, and enter the same name here. Pick something nobody would guess, because anyone who knows the topic can read it. Then Add channel, and Send test notification to check it arrives.

Further down: quiet hours, so a camera rebooting at three in the morning waits until breakfast. And maintenance windows. Planning to rewire the network on Saturday? Add a window, and the server keeps quiet while you work.

Not every device deserves the same treatment. Open a device and go to Monitoring. This ESP32 sensor runs on a battery and sleeps a lot, so it can get its own offline sensitivity, or have its notifications muted completely.

The opposite works too: set the NAS or the camera to instant, because you want to know right away when they drop.

That way the important devices alert quickly, and the flaky ones don't train you to ignore alerts.

Sometimes a device is online, but the thing you care about isn't. The NAS answers, but its web interface hangs. Under Checks and monitors, add a monitor for a specific service: a ping, a port, a web page that must return a certain text, an SNMP value or a database.

One option here saves a lot of noise: the parent. If the switch in the garage dies, everything behind it goes offline. Give the checks behind it the switch as their parent, and they stay quiet while the switch is down. For devices, the server works this out by itself: when a switch or access point fails, you get one alert about it, not one for every device behind it.

Everything that's wrong right now ends up in the Alarms view, the most serious first. Each alarm shows since when it's active, and a thirty day trend. Mute silences the notifications for one alarm while you fix it, and the alarm stays in place until the device or check is fine again.

Afterwards, Reports shows how often each device was offline and for how long, day by day. That tells you whether it was a one-off or a pattern, like a camera that drops out every night when the lights switch off.

One last thing. If the server itself loses power, it can't tell you. That's what the external watchdog is for. Sign up for a free heartbeat service like Healthchecks dot io, paste its address here, and the server checks in every minute. When the check-ins stop, that service alerts you.

To sum up: switch on the offline alert, use standard sensitivity and a few minutes of delay, and send it to your phone with ntfy or by e-mail. Tune single devices up or down, add checks for services, set a parent so one failure gives one alert, and let a watchdog keep an eye on the server itself. DeviceShelf Server is linked in the description.