NVR offline: why it happens and how to be told

An NVR shows offline when the app or monitoring software can no longer reach it. The usual causes are power, a cable or switch, a changed address after a power cut, the internet line, a replaced router, or the maker's P2P service.

A recorder that has lost power or its network cannot send a message about it. The check has to come from something else on the site's network that is still running, or from a service that notices the site has gone quiet.

Offline means nothing reached it

"Offline" in an app or in monitoring software means the connection failed. It does not say where. The recorder may be off, cut off from the network, at a new address, or fine and unreachable from outside.

So the first question is how far the problem reaches. If one camera is missing, the recorder is probably up. If the recorder and everything else at the site is unreachable, look at power, the router and the internet line first.

The usual causes and how to check them

CGNAT is carrier-grade NAT: the site shares a public address with many others, so nothing can be forwarded to it. Sites on 4G, 5G and Starlink are almost always behind it. A port forward cannot work there, whatever the recorder does.

After a power cut

A power cut is the most common way to find several of these causes at once. When power returns, the router, the switch and the recorder all restart, in whatever order they come up.

1. Check the recorder has started and its cameras show on its own screen.
2. Check the switch's link lights for the recorder and the cameras.
3. Check the router has its internet connection back.
4. Check the recorder's address. A recorder that takes its address from the router (DHCP) can be given a different one after a restart. Reserve its address in the router, or set a fixed address outside the router's DHCP range.
5. Check remote access from outside the site, not from a phone on the site's Wi-Fi.

A recorder cannot report its own outage

A recorder's own notifications, where it has them, are sent by the recorder. A recorder with no power, or no path to the internet, has no way to send anything. The people who should know find out when someone next looks, which may be after an incident.

So the check has to come from somewhere else. Something on the same network, still running, has to notice the recorder has stopped answering. And something outside the site has to notice when the whole site goes quiet, because then nothing at the site can say so.

How our box checks and alerts

A Relayant box sits on a LAN port of the site's router. Any device on the site's network list can be monitored, recorders included. The box checks each one every minute and reports when it stops answering.

• Three results. The device is working; it is on the network but not answering, as a hung recorder is; or it is down. A hung recorder and an unplugged one need different people.
• The strongest check that works. The box learns it per device, from ARP and ping through TCP, RTSP, ONVIF and HTTP. You can override it.
• Steady, not twitchy. A failed check is retried once. A device is marked faulty or down only after three results in a row over at least two and a half minutes, so a single missed check raises no alert.

It follows a recorder that changes address. The box follows a device by its hardware (MAC) address when its IP changes. A recorder you have connected keeps the same address and port on Relayant, so the monitoring software needs no change.

It tells a site outage from a device outage. When most of a site's devices drop at once, you get one site alert, not one per device. If the power cut or the line fault takes the box down too, the portal shows the box offline. On Relay + Alerts, an alert is raised. The box reconnects by itself once the site has power and internet and Relayant is reachable.

Our service raises these alerts. While a site cannot reach us, its box is reported offline and nothing else is raised for its devices.

Alerts go where your people look. Alerts are shown in the portal and sent by email, to Slack and Microsoft Teams channels, and to webhooks. A control room can also have each fault sent into its monitoring software as an alarm signal over SIA DC-09. Alerts for a device are held while it is muted for planned work. Sending is in the Relay + Alerts plan; on Relay, alerts appear in the portal only.

The details are on Device health and Alerts and alarm signals.

What the box does not check

The box checks that a recorder answers. Video content, picture quality and recording are not checked. That needs the device's login, which we do not hold.

Checks are gentle: one probe at a time per device and no login attempt. RTSP checks use OPTIONS only, so they use no recorder stream slot. Vendor ports that need a login, such as 8000 or 37777, are only checked for an open port.

Something in this guide is wrong or missing? Tell us: hello@relayant.net.