Most people find out what their camera system does without internet on the worst possible day for it. The storm knocks the line out, something happens in the driveway, and the next morning there’s a two-hour hole in the timeline and no clear reason why.
The whole thing is avoidable with one test that costs fifteen minutes and a walk to the router. We mentioned it briefly at the end of the off-grid post. This is the full version, with what to expect and what each result means.
The test
- Find the cable that goes from your router to the wall, the modem, or the fiber box. That’s the uplink. Not the power cable, not the cables to your switch.
- Note the time.
- Unplug the uplink and leave it out for fifteen minutes. Fifteen, not two. Some failures take a few minutes to show up because a cached login or a token is still good.
- During the fifteen minutes, from a phone or laptop on your home Wi-Fi, try each item in the table below and write down what happened.
- Plug the uplink back in. Wait five minutes.
- Go look at the recording timeline for the window you just created.
The whole point is step 6. Everything else is setup.
What to check while the cable is out
Work down this list and mark each one survived or died.
| While the internet is out… | Survives | Dies |
|---|---|---|
| Cameras still show a live picture from a device on your Wi-Fi | Video path is local | Video is routed through a vendor server even at home |
| The recorder is still writing continuous video | Storage is local | Recording depends on a cloud upload |
| Motion or object detection is still producing events | Detection runs on your hardware | Detection is done off-site |
| You can open the recorder’s own web page or app on Wi-Fi | Access is local | Login is checked against a vendor account server |
| You can scrub back to footage from ten minutes ago | Playback is local | Clips are fetched from cloud storage |
| A phone that is off your Wi-Fi gets a notification | You have a private tunnel that survived somehow | Expected. Push needs the internet, full stop |
| A phone that is off your Wi-Fi can see live video | Same as above | Expected. See below |
The last two rows are there to make a point, not to be passed. Nothing that reaches a phone outside your house can work while your house is offline. If those two “die,” your system is behaving normally.
Reading the results
Everything in the first five rows survived. Your storage, detection, viewing and access are all local. An internet outage is invisible to the recorder. The only things you lose are remote viewing and push, which come back on their own when the line does. This is where a Frigate setup with local storage and a local detector should land, and if you set it up yourself it’s worth fifteen minutes to confirm it rather than assume it.
Recording survived but detection died. The video is safe, but the system stops deciding what’s interesting. When the internet comes back you’ll have continuous footage and a gap in the event list, which means anything that happened during the outage has to be found by scrubbing. This is a common shape for camera brands that record to a local card or drive but do their person detection in the cloud.
Live view died on your own Wi-Fi. This is the one that surprises people most. The camera is twenty feet away, the phone is on the same network, and the picture won’t load, because the app’s path is phone to vendor to camera even when both ends are in the same room. Nothing you can configure fixes this. It’s a decision the vendor made.
The login page died. You can’t get into your own recorder without the vendor’s account server saying yes. Everything might still be recording behind that locked door, but during the outage you can’t check, and if the vendor’s server is the thing that’s down rather than your line, same result.
There’s a gap in the timeline after the cable went back in. This is the result that should change your weekend plans. Either recording actually stopped, or the recorder was writing to a buffer it never flushed. Check the disk, check the logs, and if you can’t explain it, treat the system as one that doesn’t record during outages, because that’s what the evidence says.
The fifteen minutes after
A few things to look at once the line is back:
- Did the cameras themselves reconnect? Some cameras reboot when they lose their NTP or cloud heartbeat and come back on a different IP if your router hands out new leases. A reserved address per camera on the router avoids that.
- Did anything need a manual login? If a device sat waiting for you to re-authenticate, it would have sat that way for the whole outage in real life.
- Are the timestamps right? A recorder that lost its clock and didn’t have a local time source can stamp an hour of footage with the wrong time, which is nearly as bad as not having it.
What about a longer outage
Fifteen minutes proves the architecture. It does not prove endurance. A multi-hour outage brings in two more questions that this test doesn’t answer.
The first is power, because the same storm that takes out the line often takes out the mains, and a recorder with no battery behind it fails the whole test for a different reason. That’s a sizing problem with real numbers in it, and it’s the next post.
The second is disk. A system that records continuously fills its drive at the same rate whether or not anyone can watch, so a long outage isn’t a storage problem on its own. But if your setup normally offloads clips to the cloud and keeps only a small local buffer, a long outage can wrap that buffer and overwrite the footage you wanted. Know which kind you have.
What we can say about push
Since this comes up every time the topic does: ViewPane is a viewer for Frigate, and the push notifications it can show you are forwarded by a small relay that runs on your own hardware. That relay is local, and it keeps running during an outage. But the message it produces still has to leave your house and travel through Apple’s or Google’s push service to reach a phone that isn’t on your Wi-Fi. No internet, no delivery. A send that fails can be retried while the same event is still updating, but events that started and finished during the gap are not replayed once the line comes back.
That is not a ViewPane limitation. It is how push works for every camera app, cloud or local. Treat notifications as best-effort on any system, and treat the recorder as the thing that actually has to survive the outage, because it is the only part of the system that can.
The one-line version
Unplug the uplink for fifteen minutes and go look at the timeline afterwards. If there’s no gap and you could watch live video the whole time from your couch, your setup is genuinely local. If there’s a gap, you’ve just learned something important on a day when it didn’t cost you anything.