Skip to Content

Bundle data (via JavaScript)

By Thorsten Schneider

Bundle sensor data before storing it in HDFS


You will learn

During this tutorial, you will learn how you can “bundle” (in the sense that not each single sensor record creates a new file) the sensor data before you store it in HDFS. You will use a JavaScript Operator for this.

Step 1: Add JavaScript Operator

Open the pipeline which you have created during the previous tutorials (test.myFirstPipeline) in the modelling environment http://localhost:8090.

Remove the connection between the Kafka Consumer operator and the HDFS Producer operator.

Add a ToString Converter operator to the pipeline by drag & drop. Also add a JavaScript Operator to the pipeline by drag & drop.

Connect the message port of the Kafka Consumer operator to the ininterface port of the ToString Converter operator.
Connect the outstring port of the ToString Converter operator to the input port of the JavaScript Operator.
Connect the output port of the JavaScript Operator to the inFile port of the HDFS Producer operator.

Step 2: Create JavaScript extension

Press the Script button to display the JavaScript snippet which is executed by the JavaScript Operator. The JavaScript snippet opens in a new tab.


Currently the JavaScript snippet creates an incremental counter every time it receives data via the input port and sends the counter to the output port.

Replace the code with the following snippet to ensure that “bundles” of 30 sensor records are sent to the output port.

var counter = 0;
var bundle = "";


function onInput(ctx,s) {
    bundle = bundle + s;

    if(counter==30) {
      counter = 0;
      bundle = "";

Close the tab for the JavaScript snippet. Afterwards press the Save button.

Step 3: Execute the data pipeline

Before you execute the pipeline, delete the contents from the /tmp/hdfsManager directory. Thereto log into the running container.

docker exec -it datahub /bin/bash

Then enter the following command.

hdfs dfs -rm /tmp/hdfsManager/*

You can check the deletion via http://localhost:50070.

Press the Run button to execute the pipeline.

When the Status tab indicates that the pipeline is running, use the context menu Open UI of the Terminal operator to see the generated sensor data.

Open http://localhost:50070 and display the /tmp/hdfsManager directory. You see that the system does not create a file for each single sensor record, but only for each 30 sensor records.

Stop the pipeline by pressing the Stop button.

Step 3: Check the created files in HDFS

To check the content of the created files, log into the running container and enter the following (to see, for example, the content of file test_1.txt ).

hdfs dfs -cat /tmp/hdfsManager/test_1.txt

Attention: You might notice that the first line of file test_1.txt does not necessarily have the counter 0 (that is the first column of the file and in the above screenshot it has the counter 123) as one might expect when looking at the JavaScript snippet which is executed by the Data Generator. The reason for this can be that the Kafka Producer had send a message to Kafka, but then you most likely stopped the pipeline before the Kafka Consumer consumed the message. When you afterwards restarted the pipeline, the Kafka Consumer first of all processed this “previous” message. For the sake of this tutorial you do not have to bother about this behavior.

Next Steps

Store sensor data in SAP Vora

Updated 12/19/2017

Time to Complete

15 Min




Back to top