Your hub went offline. Was it the power or the internet?

The phone buzzes at twenty to four: the hub has gone offline. You are ninety minutes away, the dog is in the rig, and the message tells you precisely one thing — the thing you were relying on has stopped talking to you. It does not tell you whether the rig still has power.

That one alert covers two situations that could hardly be more different. If the internet went, the hub is still sitting there reading every sensor and writing all of it to its own card, and the whole record arrives the moment the link returns. You lost the alerts, not the data. If the power went, nothing was recorded, the gap in the history is a real gap, and if the hub shares a supply with the thing you were watching, the gap is the emergency.

Same alert for both. Our servers cannot tell them apart and do not pretend to. Four things can. Two of them cost nothing, and one of those two is definitive.

Why the two look the same

A hub cannot report its own death. Whatever kills it takes away the thing it would have used to tell you, so both failures reach us as the same event: check-ins that stop arriving. The offline alert is our servers noticing an absence rather than the hub reporting a problem, which makes it the only alert on the site you never configure — no threshold, every plan, fires on its own — and the only one that is ambiguous by construction.

There is a grace period first, deliberately measured in minutes rather than seconds, because a hub rebooting after a firmware update must not page you. And it is an ordinary alert rather than a safety alarm, so it observes your quiet hours like everything else.

Every way to tell them apart

a. The backlog fills the gap, or it does not

This is the definitive one and it costs nothing. When the hub reconnects, open the chart of a sensor that reports every minute and was reliable before the gap — a Govee H5075, a RuuviTag, the voltage line from a Victron shunt. If readings fill the whole gap, the hub had power throughout and you lost the internet. If there is a flat hole running from the offline alert to the reconnect, the hub had no power.

The backlog is complete: nothing expires, nothing is dropped, no plan tier limits how far back it goes. Speed is the only ceiling, at 1,000 readings a minute — six sensors on 60-second intervals across a full day is 8,640 readings, drained in under nine minutes. The catch is the tense. This method only answers in the past, and if the hub never comes back no backlog ever arrives. That silence is itself the answer, and the answer is power.

b. The shore power monitor’s last frames

If you are on hookups with a Hughes Power Watchdog, read the minutes before the hub went quiet. It polls every 30 seconds, so a pedestal event and the hub’s silence land within one poll of each other. Normal voltage right up to the last frame means shore power was fine when the hub vanished — the internet, a switched outlet, or a hub somebody moved to a different outlet. The Watchdog dropping out in the same minute means the pedestal took its radio with it, and if your hub is on an AC outlet that is why the hub died too. An error code just before means its EPO caught a fault and cut the outlet deliberately, which is the shore power article, not this one.

The weakness is that the Watchdog is a connection-based device: a phone running the Hughes app can hold that connection and leave the hub reading it as absent while the pedestal is perfectly healthy. And it only ever speaks about the pedestal. It has nothing to say about a hub running on 12 volts.

c. The house battery’s last reading

If the hub is wired to 12 volts — and it should be — the house battery is its supply, and a battery monitor is already watching that supply. A SmartShunt reads every 60 seconds, a Lancol every 120, so you get at least one honest reading before the lights go out. That is enough: a last reading at or below your critical floor says the bank ran down and took the hub with it, and a healthy one rules the battery out.

On lead-acid, voltage answers this well. On lithium it barely answers at all: a LiFePO4 bank sits around 13.2 volts from ninety per cent down to about twenty and then falls off a cliff, so a comfortable-looking number proves nothing. On lithium it is a shunt’s counted percentage or it is a guess, which is the whole argument for a shunt compressed into one sentence.

d. The clock it happens at

Free, and better than it sounds. An offline alert at roughly the same time every night, followed by a complete backlog each morning, is a captive portal logging the hub out on a 24-hour timer. A hub cannot click “I accept”. A campground that reboots its access point on a schedule looks identical and wants the same fix, so there is no point telling those two apart — once it repeats on a clock, stop diagnosing and buy the travel router that holds the session for you. An alert at a random hour is the one to read with the methods above.

One pattern gets its own article, because it happens on every trip and hides something worth finding: the offline alert that arrives as you pull out of a site. Usually that is only the campground Wi-Fi falling behind you. Occasionally it is a hub on an outlet switched with the ignition, which makes every drive a blind gap.

The combinations

As everywhere else in the cookbook, one reading says something happened and two say what.

• Offline, the Watchdog dropped in the same minute, hub on an AC outlet — the pedestal went and took the hub with it. Really a note about your own install: on 12 volts that same event arrives as a shore power alert from a hub that is still watching.
• Offline, the Watchdog normal to its last frame, battery healthy — the internet. The hub is logging, and the record is waiting for you.
• Offline at the same time nightly, complete backlog each time — a portal. One router fixes it permanently.
• Offline at a random hour, complete backlog, and the Watchdog’s own readings stop in the same minute — the campground lost power and network together. A 12-volt hub rode it out; an AC-powered one would have gone dark and told you nothing. An error code instead of a gap is a different story: the EPO tripped on a fault it caught, which is the pedestal article.
• Offline, and it never comes back — power, and no backlog is coming. Everything from the gap is genuinely gone.

What none of this tells you

Nothing here works live. Every method above is read from the last minutes before the silence or the first minutes after it, because in between there is no hub to ask. If you need to know at four and the hub went quiet at two, you are inferring rather than measuring.

So make a habit of the other half: after every reconnect, open the charts of the two or three sensors you actually care about and read the gap. It is the only way to find out what you missed. The backlog restores the record without restoring the moment.

Design the ambiguity out instead of diagnosing it

All of the above is detective work after the fact, and most of it you can avoid with wiring, for nothing.

• Hub on 12 volts, unswitched. A shore power loss then stops being a hub power loss, and arrives as a shore power alert from a hub that is still watching.
• Your own router on the same circuit. Hub and router share a fate, and “offline” narrows to one meaning instead of three.
• A battery monitor on the bank. That last remaining meaning then gets a warning hours ahead of it — the only warning on this page that arrives before the event rather than after.

For a rig that sits somewhere with no Wi-Fi at all, the cellular backup device is sold separately and is the only way the hub reaches you from a storage lot. Power and connectivity has every combination of the three and what each one survives. Get that far and the offline alert stops being a question. It becomes what it should have been all along: rare, and specific.

Where to get one

The devices that make an offline hub readable, in the order they earn their place. Everything else on this page is a wiring decision and a habit. The compatibility database has the rest.