The Bridge and Broker Sensors

The Bridge and Broker Sensors

Device Sentinel watches every device in your home, and it also watches what those devices depend on: the coordinator they talk through, and the MQTT broker underneath it. When one of those stops, every device behind it goes quiet at once, and none of them is at fault.

These sensors are where you see that. Where a coordinator or broker is found, its sensor is switched on for you.

Bridge Sensors

One sensor for each coordinator Device Sentinel finds: Bridge: Zigbee2MQTT, Bridge: ZHA, or both on a house running both.

Shared states:

• running: the coordinator is up. This is what you see almost all the time.
• down: the coordinator has stopped, or has gone quiet long enough that its state can no longer be read.
• unknown: nothing has been heard from it yet, which you may see briefly after a restart.

Bridge: Zigbee2MQTT adds a fourth state, binding, which shows while a pairing window is open, and an availability_enabled attribute reading whether Zigbee2MQTT's own availability feature is switched on. If it reads false, turning it on adds the second opinion described on Zigbee2MQTT. Its other attributes are the base topic it reads and when the coordinator was last heard from.

Bridge: ZHA has no binding state, because ZHA publishes no pairing window that can be relied on. When you work on a ZHA device by hand, use Maintenance Mode instead.

Broker: MQTT

A coordinator cannot report that its own transport has failed, so the broker gets a sensor of its own. It reads running, down, or unknown.

Its attributes carry when the broker started, its uptime, when it was last heard from, the interval it publishes on, and the topic.

Where there is no MQTT, or where a broker publishes no statistics, the sensor reads unknown and nothing else changes. On a house with no coordinator stack the sensor is created but left switched off, so it does not sit at unknown forever.

What These States Do for You

A stopped coordinator is one row, not sixty. While a coordinator is down, the problem list, the card and your phone carry one line naming it and how many devices sit behind it, rather than a problem per casualty. Everything clears when it returns. The outage is recorded with its duration, so a mesh-wide incident reads as one event rather than sixty separate mysteries.

Silences the outage caused are not learned. A gap that spans a coordinator or broker outage measures the outage, not the device, so it is set aside. Without that, one bad night would widen every device's freeze window for a fortnight. The episode report marks those gaps no (bridge down).

A re-pair is not mistaken for the device's own behavior. On Zigbee2MQTT, a device that returns while the pairing window is open is treated as your hand rather than as a slow rhythm, and the episode report marks it no (pairing). On ZHA, Maintenance Mode does the same job.

A broker outranks a bridge, because a broker that is down takes every bridge with it.

Turning Them Off

A sensor is a window onto a state, not where the reading happens. Device Sentinel reads these states whether or not the sensors are switched on, so turning one off costs you the view and nothing else.

See Also

• The Problem List, for what a coordinator outage looks like while it is running.
• The Diagnostic Reports, for the episode vocabulary these states produce: no (bridge down), no (pairing).
• Maintenance Mode, for telling Device Sentinel that a recovery was your own hand.