Live data

A query can stream: instead of answering once, it keeps sending rows as they happen. Grafana does this through Grafana Live, its websocket, and the map draws whatever arrives: vehicles that move as they report, sensors that update, events as they come in. No dashboard refresh is involved.

Two common sources:

• Grafana Live push. Anything that can make an HTTP request posts rows to Grafana in InfluxDB line protocol, and the built-in -- Grafana -- datasource reads them with the Live measurements query type.
• MQTT, through Grafana's MQTT datasource: the datasource subscribes to a topic and streams every message.

Any other datasource that streams works the same way.

What the map does with a stream

While a query streams	What the map does
new rows arrive with every message	it updates at most once per Live update interval (1 s by default), with the newest rows; a hidden tab takes no updates and catches up when shown
the data's last instant moves forward	the trip playhead goes to it, and a time window that is not driven by the dashboard range slides to end there, with the same width (a window over the whole data keeps covering it)
you move the playhead or the window	it stays where you put it; bring it back to the end and it follows again
a series comes with labels (tags)	each label becomes a column and each series a row per instant, so bus,id=b07 latitude=…,longitude=… reads as time, id, latitude, longitude

With Time range sync on Dashboard drives map (the default) the window follows the dashboard range, which Grafana moves along with the stream; the map only moves the trip playhead.

Each update rebuilds the query's dataset in the map. At one update a second that costs little; a datasource that sends several messages a second is the reason the interval exists.

Grafana Live push

Post line protocol to /api/live/push/<stream> with a token that can edit (a service account with the Editor role):

curl -X POST "$GRAFANA/api/live/push/fleet" \
  -H "Authorization: Bearer $TOKEN" \
  --data-binary 'bus id="b07",latitude=-0.1807,longitude=-78.4678,speed=31.5,heading=90'

Each measurement is a channel, stream/<stream>/<measurement> — here stream/fleet/bus. In the panel, pick the -- Grafana -- datasource, Live measurements, and that channel.

The vehicle's id can be a field (id="b07", a string) or a tag (bus,id=b07 …): the map reads both as a column id. Grafana keeps a streaming query's newest rows in the browser, as many as the query's Max data points: about the panel's width in pixels unless set, so a thousand or so, which is 50 seconds of 20 vehicles reporting every second. Raise it under the query options for longer trails, or for a busier feed.

MQTT

Point the MQTT datasource at the broker (tcp://, tls://, ws://, or wss:// as HSL below) and query a topic; + and # wildcards work. Each message becomes a row. A JSON payload arrives as one column holding the object: add the Extract fields transformation (source: that column, format: JSON) to turn its keys into columns the map can place.

Helsinki's public transport (HSL) publishes every vehicle's position once a second, openly:

Setting	Value
Broker	wss://mqtt.hsl.fi:443/
Topic	/hfp/v2/journey/ongoing/vp/bus/# (tram, train, metro, or #)
Transformation	Extract fields, source VP, format JSON, replace all fields
Columns	lat, long, veh (vehicle number), oper (operator), spd, hdg

A vehicle number is unique within its operator; for a trip per vehicle, map Field mapping → Trip ID to veh (or to a column that joins oper and veh).

See it live at pacha.ubica.dev/d/hsl-live: the same feed on a Kepler Plus Fleet layer, one kind of vehicle at a time, keeping 20000 rows (Max data points, above) for hundreds of vehicles a second.

Options

Under Live data:

Option	Default	What it does
Follow live data	on	keeps the playhead and the window on the newest rows, as above
Live update interval (ms)	1000	the least time between two updates of the map while a query streams

---

Written against kepler.gl 3.3.0-alpha.20. See Differences from kepler.gl for what changes inside a Grafana panel.