EMQX

Use Case: Predictive maintenance

Logistics use case (predictive maintenance)

You want near-real-time telemetry from delivery trucks.

Each truck publishes sensor telemetry to an MQTT topic over GSM.

• EMQX is the MQTT broker.

• Pentaho Data Integration (PDI) subscribes and processes messages.

You will:

• publish telemetry into EMQX (producer transformation)

• consume telemetry in PDI (consumer + child transformation)

• write minor and major alerts to output files

Logistics

Before you begin

Make sure EMQX is running and reachable:

• MQTT broker (TCP): localhost:1883

• EMQX Dashboard (web): http://localhost:18083

Default dashboard credentials are often:

• Username: admin

• Password: Password123

If you need to set it up first, see EMQX setup.

Make sure the lab files exist:

• ~/Streaming/emqx/solution/tr_mqtt_producer.ktr

• ~/Streaming/emqx/solution/tr_mqtt_consumer.ktr

• ~/Streaming/emqx/solution/tr_process_sensor_data.ktr

EMQX: The Unified MQTT Platform for AI and IoT Data Streamingwww.emqx.com

1

Start Pentaho Data Integration

Launch Data Integration:

macOS / Linux

cd
cd ~/Pentaho/design-tools/data-integration
./spoon.sh

Windows

2

Publish telemetry to EMQX (MQTT Producer)

Open:

~/Streaming/emqx/solution/tr_mqtt_producer.ktr

MQTT Producer

This transformation:

• generates data for vehicle_id = 111 every 5 seconds

• adds a timestamp

• builds a message payload

• publishes the payload to EMQX using MQTT Producer

Double-click MQTT Producer and confirm these settings:

• Connection: points to your broker (for example tcp://localhost:1883)

• Client ID: unique on the broker

• Topic: note the value. You will reuse it in the consumer.

• QoS: choose based on delivery requirements

• Message field: message

MQTT Producer

Run the transformation.

Validate in EMQX Dashboard:

1. Open http://localhost:18083.

2. Go to Clients. Confirm the producer client is connected.

3. Go to Topics. Confirm message rate on your topic.

Reference: QoS and session settings

QoS

• 0 (at most once): lowest latency. Messages can be lost.

• 1 (at least once): can deliver duplicates.

• 2 (exactly once): highest overhead.

Clean session

• True: broker does not persist state for this client.

• False: broker persists subscriptions and queued QoS 1/2 messages.

3

Consume and process telemetry (MQTT Consumer)

Open:

~/Streaming/emqx/solution/tr_mqtt_consumer.ktr

MQTT Consumer

Double-click MQTT Consumer and confirm:

• Topic matches the producer topic.

• Child transformation points to:

  ${Internal.Entry.Current.Directory}/tr_process_sensor_data.ktr

• Batch has at least one trigger set:

  • Duration (ms) > 0, or

  • Number of records > 0

MQTT Consumer

Run the transformation.

Validate in EMQX Dashboard:

• Go to Clients. Confirm the consumer client is connected.

• Go to Subscriptions. Confirm the client is subscribed to the topic.

Reference: batch and backpressure

PDI runs the child transformation when either threshold is met:

• Duration (ms): time window for collecting messages

• Number of records: message count for collecting messages

Backpressure controls:

• Message prefetch limit: max messages PDI queues in memory

• Maximum concurrent batches: increases throughput and resource use

4

Inspect results

Open the child transformation:

~/Streaming/emqx/solution/tr_process_sensor_data.ktr

Process sensor data

This transformation:

• pulls message records from the stream

• adds a timestamp

• resolves sensor_type

• filters minor and major alerts

• aggregates alerts for reporting

• appends results to output files

Open the output files and confirm new rows append:

• ~/Streaming/emqx/output/major_alert.txt

• ~/Streaming/emqx/output/minor_alert.txt

major_alert

minor_alert

If you see no output, check these first:

• producer and consumer use the same topic

• EMQX shows both clients as connected

• the consumer batch has Duration (ms) or Number of records set to > 0

• the child transformation starts with Get records from stream