Debug Apache Spark applications on an HDInsight cluster with Azure Toolkit for IntelliJ through SSH

This article provides step-by-step guidance on how to use HDInsight Tools in Azure Toolkit for IntelliJ to debug applications remotely on an HDInsight cluster. To debug your project, you can also view the Debug HDInsight Spark applications with Azure Toolkit for IntelliJ video.


Learn how to perform local run and debugging

Scenario 1: Create a Spark Scala application

  1. Start IntelliJ IDEA, and then create a project. In the New Project dialog box, do the following:

    a. Select Azure Spark/HDInsight.

    b. Select a Java or Scala template based on your preference. Select between the following options:

    • Spark Project (Java)

    • Spark Project (Scala)

    • Spark Project with Samples (Scala)

    • Spark Project with Failure Task Debugging Samples (Preview) (Scala)

      This example uses a Spark Project with Samples (Scala) template.

    c. In the Build tool list, select either of the following, according to your need:

    • Maven, for Scala project-creation wizard support

    • SBT, for managing the dependencies and building for the Scala project

      Create a debug project

    d. Select Next.

  2. In the next New Project window, do the following:

    Select the Spark SDK

    a. Enter a project name and project location.

    b. In the Project SDK drop-down list, select Java 1.8 for Spark 2.x cluster or select Java 1.7 for Spark 1.x cluster.

    c. In the Spark Version drop-down list, the Scala project creation wizard integrates the correct version for Spark SDK and Scala SDK. If the spark cluster version is earlier than 2.0, select Spark 1.x. Otherwise, select Spark 2.x. This example uses Spark 2.0.2 (Scala 2.11.8).

    d. Select Finish.

  3. Select src > main > scala to open your code in the project. This example uses the SparkCore_wasbloTest script.

Prerequisite for Windows

While you're running the local Spark Scala application on a Windows computer, you might get an exception, as explained in SPARK-2356. The exception occurs because WinUtils.exe is missing on Windows.

To resolve this error, download the executable to a location such as C:\WinUtils\bin. Then, add the environment variable HADOOP_HOME, and set the value of the variable to C:\WinUtils.

Scenario 2: Perform local run

  1. Open the SparkCore_wasbloTest script, right-click the script editor, and then select the option Run '[Spark Job]XXX' to perform local run.

  2. Once local run completed, you can see the output file save to your current project explorer data > default.

    Local run result

  3. Our tools have set the default local run configuration automatically when you perform the local run and local debug. Open the configuration [Spark on HDInsight] XXX on the upper right corner, you can see the [Spark on HDInsight]XXX already created under Apache Spark on HDInsight. Switch to Locally Run tab.

    Local run configuration

    • Environment variables: If you already set the system environment variable HADOOP_HOME to C:\WinUtils, it can auto detect that no need to manually add.
    • WinUtils.exe Location: If you have not set the system environment variable, you can find the location by clicking its button.
    • Just choose either of two options and, they are not needed on MacOS and Linux.
  4. You can also set the configuration manually before performing local run and local debug. In the preceding screenshot, select the plus sign (+). Then select the Apache Spark on HDInsight option. Enter information for Name, Main class name to save, then click the local run button.

Scenario 3: Perform local debugging

  1. Open the SparkCore_wasbloTest script, set breakpoints.
  2. Right-click the script editor, and then select the option Debug '[Spark on HDInsight]XXX' to perform local debugging.

Learn how to perform remote run and debugging

Scenario 1: Perform remote run

  1. To access the Edit Configurations menu, select the icon in the upper-right corner. From this menu, you can create or edit the configurations for remote debugging.

    Edit configurations

  2. In the Run/Debug Configurations dialog box, select the plus sign (+). Then select the Apache Spark on HDInsight option.

    Add new configuration

  3. Switch to Remotely Run in Cluster tab. Enter information for Name, Spark cluster, and Main class name. Then Click Advanced configuration (Remote Debugging). Our tools support debug with Executors. The numExectors, the default value is 5. You'd better not set higher than 3.

    Run debug configurations

  4. In the Advanced Configuration (Remote Debugging) part, select Enable Spark remote debug. Enter the SSH username, and then enter a password or use a private key file. If you want to perform remote debug, you need to set it. There is no need to set it if you just want to use remote run.

    Enable Spark remote debug

  5. The configuration is now saved with the name you provided. To view the configuration details, select the configuration name. To make changes, select Edit Configurations.

  6. After you complete the configurations settings, you can run the project against the remote cluster or perform remote debugging.

    Remote run button

  7. Click the Disconnect button that the submission logs not appear in the left panel. However, it is still running on the backend.

    Remote run result

Scenario 2: Perform remote debugging

  1. Set up breaking points, and then Click the Remote debug icon. The difference with remote submission is that SSH username/password need to be configured.

    Select the debug icon

  2. When the program execution reaches the breaking point, you see a Driver tab and two Executor tabs in the Debugger pane. Select the Resume Program icon to continue running the code, which then reaches the next breakpoint. You need to switch to the correct Executor tab to find the target executor to debug. You can view the execution logs on the corresponding Console tab.

    Debugging tab

Scenario 3: Perform remote debugging and bug fixing

  1. Set up two breaking points, and then select the Debug icon to start the remote debugging process.

  2. The code stops at the first breaking point, and the parameter and variable information are shown in the Variables pane.

  3. Select the Resume Program icon to continue. The code stops at the second point. The exception is caught as expected.

    Throw error

  4. Select the Resume Program icon again. The HDInsight Spark Submission window displays a "job run failed" error.

    Error submission

  5. To dynamically update the variable value by using the IntelliJ debugging capability, select Debug again. The Variables pane appears again.

  6. Right-click the target on the Debug tab, and then select Set Value. Next, enter a new value for the variable. Then select Enter to save the value.

    Set value

  7. Select the Resume Program icon to continue to run the program. This time, no exception is caught. You can see that the project runs successfully without any exceptions.

    Debug without exception

Next steps



Create and run applications

Tools and extensions

Manage resources