Building Spark .NET on Windows

Open Issues:

Allow users to choose which .NET framework to build for
Building through Visual Studio Code
Building fully automatically through .NET Core CLI

Pre-requisites:

If you already have all the pre-requisites, skip to the build steps below.

Download and install the .NET 6 SDK - installing the SDK will add the dotnet toolchain to your path.
Install Visual Studio 2019 (Version 16.4 or later). The Community version is completely free. When configuring your installation, include these components at minimum:
- .NET desktop development
  - All Required Components
    - .NET Framework 4.6.1 Development Tools
- .NET 6 cross-platform development
  - All Required Components
Install Java 1.8
- Select the appropriate version for your operating system e.g., jdk-8u201-windows-x64.exe for Win x64 machine.
- Install using the installer and verify you are able to run java from your command-line
Install Apache Maven 3.6.3+
- Download Apache Maven 3.6.3
- Extract to a local directory e.g., c:\bin\apache-maven-3.6.3\
- Add Apache Maven to your PATH environment variable e.g., c:\bin\apache-maven-3.6.3\bin
- Verify you are able to run mvn from your command-line
Install Apache Spark 2.3+
- Download Apache Spark 2.3+ and extract it into a local folder (e.g., c:\bin\spark-2.3.2-bin-hadoop2.7\) using 7-zip.
- Add Apache Spark to your PATH environment variable e.g., c:\bin\spark-2.3.2-bin-hadoop2.7\bin
- Add a new environment variable SPARK_HOME e.g., C:\bin\spark-2.3.2-bin-hadoop2.7\
- Verify you are able to run spark-shell from your command-line
  📙 Click to see sample console output
```
Welcome to
      ____              __
     / __/__  ___ _____/ /__
    _\ \/ _ \/ _ `/ __/  '_/
   /___/ .__/\_,_/_/ /_/\_\   version 2.3.2
      /_/

Using Scala version 2.11.8 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_201)
Type in expressions to have them evaluated.
Type :help for more information.

scala> sc
res0: org.apache.spark.SparkContext = org.apache.spark.SparkContext@6eaa6b0c
```
Install WinUtils
- Download winutils.exe binary from WinUtils repository. You should select the version of Hadoop the Spark distribution was compiled with, e.g. use hadoop-2.7.1 for Spark 2.3.2.
- Save winutils.exe binary to a directory of your choice e.g., c:\hadoop\bin
- Set HADOOP_HOME to reflect the directory with winutils.exe (without bin). For instance, using command-line:
```
set HADOOP_HOME=c:\hadoop
```
- Set PATH environment variable to include %HADOOP_HOME%\bin. For instance, using command-line:
```
set PATH=%HADOOP_HOME%\bin;%PATH%
```

Please make sure you are able to run dotnet, java, mvn, spark-shell from your command-line before you move to the next section. Feel there is a better way? Please open an issue and feel free to contribute.

Note: A new instance of the command-line may be required if any environment variables were updated.

Building

For the rest of the section, it is assumed that you have cloned Spark .NET repo into your machine e.g., c:\github\dotnet-spark\

git clone https://github.com/dotnet/spark.git c:\github\dotnet-spark

Building Spark .NET Scala Extensions Layer

When you submit a .NET application, Spark .NET has the necessary logic written in Scala that inform Apache Spark how to handle your requests (e.g., request to create a new Spark Session, request to transfer data from .NET side to JVM side etc.). This logic can be found in the Spark .NET Scala Source Code.

Regardless of whether you are using .NET Framework or .NET 6, you will need to build the Spark .NET Scala extension layer. This is easy to do:

cd src\scala
mvn clean package

You should see JARs created for the supported Spark versions:

microsoft-spark-2-3\target\microsoft-spark-2-3_2.11-<version>.jar
microsoft-spark-2-4\target\microsoft-spark-2-4_2.11-<version>.jar
microsoft-spark-3-0\target\microsoft-spark-3-0_2.12-<version>.jar

Building .NET Samples Application

Using Visual Studio for .NET Framework

Open src\csharp\Microsoft.Spark.sln in Visual Studio and build the Microsoft.Spark.CSharp.Examples project under the examples folder (this will in turn build the .NET bindings project as well). If you want, you can write your own code in the Microsoft.Spark.Examples project:

  // Instantiate a session
  var spark = SparkSession
      .Builder()
      .AppName("Hello Spark!")
      .GetOrCreate();

  var df = spark.Read().Json(args[0]);

  // Print schema
  df.PrintSchema();

  // Apply a filter and show results
  df.Filter(df["age"] > 21).Show();

Once the build is successfuly, you will see the appropriate binaries produced in the output directory.

📙 Click to see sample console output

      Directory: C:\github\dotnet-spark\artifacts\bin\Microsoft.Spark.CSharp.Examples\Debug\net461


  Mode                LastWriteTime         Length Name
  ----                -------------         ------ ----
  -a----         3/6/2019  12:18 AM         125440 Apache.Arrow.dll
  -a----        3/16/2019  12:00 AM          13824 Microsoft.Spark.CSharp.Examples.exe
  -a----        3/16/2019  12:00 AM          19423 Microsoft.Spark.CSharp.Examples.exe.config
  -a----        3/16/2019  12:00 AM           2720 Microsoft.Spark.CSharp.Examples.pdb
  -a----        3/16/2019  12:00 AM         143360 Microsoft.Spark.dll
  -a----        3/16/2019  12:00 AM          63388 Microsoft.Spark.pdb
  -a----        3/16/2019  12:00 AM          34304 Microsoft.Spark.Worker.exe
  -a----        3/16/2019  12:00 AM          19423 Microsoft.Spark.Worker.exe.config
  -a----        3/16/2019  12:00 AM          11900 Microsoft.Spark.Worker.pdb
  -a----        3/16/2019  12:00 AM          23552 Microsoft.Spark.Worker.xml
  -a----        3/16/2019  12:00 AM         332363 Microsoft.Spark.xml
  ------------------------------------------- More framework files -------------------------------------

Using .NET CLI for .NET 6

Note: We are currently working on automating .NET 6 builds for Spark .NET. Until then, we appreciate your patience in performing some of the steps manually.

Build the Worker

cd C:\github\dotnet-spark\src\csharp\Microsoft.Spark.Worker\
dotnet publish -f net6.0 -r win-x64

📙 Click to see sample console output

PS C:\github\dotnet-spark\src\csharp\Microsoft.Spark.Worker> dotnet publish -f net6.0 -r win-x64
Microsoft (R) Build Engine version 16.0.462+g62fb89029d for .NET Core
Copyright (C) Microsoft Corporation. All rights reserved.

  Restore completed in 299.95 ms for C:\github\dotnet-spark\src\csharp\Microsoft.Spark\Microsoft.Spark.csproj.
  Restore completed in 306.62 ms for C:\github\dotnet-spark\src\csharp\Microsoft.Spark.Worker\Microsoft.Spark.Worker.csproj.
  Microsoft.Spark -> C:\github\dotnet-spark\artifacts\bin\Microsoft.Spark\Debug\netstandard2.0\Microsoft.Spark.dll
  Microsoft.Spark.Worker -> C:\github\dotnet-spark\artifacts\bin\Microsoft.Spark.Worker\Debug\net6.0\win-x64\Microsoft.Spark.Worker.dll
  Microsoft.Spark.Worker -> C:\github\dotnet-spark\artifacts\bin\Microsoft.Spark.Worker\Debug\net6.0\win-x64\publish\

Build the Samples

cd C:\github\dotnet-spark\examples\Microsoft.Spark.CSharp.Examples\
dotnet publish -f net6.0 -r win-x64

📙 Click to see sample console output

PS C:\github\dotnet-spark\examples\Microsoft.Spark.CSharp.Examples> dotnet publish -f net6.0 -r win10-x64
Microsoft (R) Build Engine version 16.0.462+g62fb89029d for .NET Core
Copyright (C) Microsoft Corporation. All rights reserved.

  Restore completed in 44.22 ms for C:\github\dotnet-spark\src\csharp\Microsoft.Spark\Microsoft.Spark.csproj.
  Restore completed in 336.94 ms for C:\github\dotnet-spark\examples\Microsoft.Spark.CSharp.Examples\Microsoft.Spark.CSharp.Examples.csproj.
  Microsoft.Spark -> C:\github\dotnet-spark\artifacts\bin\Microsoft.Spark\Debug\netstandard2.0\Microsoft.Spark.dll
  Microsoft.Spark.CSharp.Examples -> C:\github\dotnet-spark\artifacts\bin\Microsoft.Spark.CSharp.Examples\Debug\net6.0\win-x64\Microsoft.Spark.CSharp.Examples.dll
  Microsoft.Spark.CSharp.Examples -> C:\github\dotnet-spark\artifacts\bin\Microsoft.Spark.CSharp.Examples\Debug\net6.0\win-x64\publish\

Run Samples

Once you build the samples, running them will be through spark-submit regardless of whether you are targeting .NET Framework or .NET 6 apps. Make sure you have followed the pre-requisites section and installed Apache Spark.

Set the DOTNET_WORKER_DIR or PATH environment variable to include the path where the Microsoft.Spark.Worker binary has been generated (e.g., c:\github\dotnet\spark\artifacts\bin\Microsoft.Spark.Worker\Debug\net461 for .NET Framework, c:\github\dotnet-spark\artifacts\bin\Microsoft.Spark.Worker\Debug\net6.0\win-x64\publish for .NET 6)
Open Powershell and go to the directory where your app binary has been generated (e.g., c:\github\dotnet\spark\artifacts\bin\Microsoft.Spark.CSharp.Examples\Debug\net461 for .NET Framework, c:\github\dotnet-spark\artifacts\bin\Microsoft.Spark.CSharp.Examples\Debug\net6.0\win1-x64\publish for .NET 6)

Running your app follows the basic structure:

spark-submit.cmd `
  [--jars <any-jars-your-app-is-dependent-on>] `
  --class org.apache.spark.deploy.dotnet.DotnetRunner `
  --master local `
  <path-to-microsoft-spark-jar> `
  <path-to-your-app-exe> <argument(s)-to-your-app>

Here are some examples you can run:

Microsoft.Spark.Examples.Sql.Batch.Basic

spark-submit.cmd `
--class org.apache.spark.deploy.dotnet.DotnetRunner `
--master local `
C:\github\dotnet-spark\src\scala\microsoft-spark-<version>\target\microsoft-spark-<version>.jar `
Microsoft.Spark.CSharp.Examples.exe Sql.Batch.Basic %SPARK_HOME%\examples\src\main\resources\people.json

Microsoft.Spark.Examples.Sql.Streaming.StructuredNetworkWordCount

spark-submit.cmd `
--class org.apache.spark.deploy.dotnet.DotnetRunner `
--master local `
C:\github\dotnet-spark\src\scala\microsoft-spark-<version>\target\microsoft-spark-<version>.jar `
Microsoft.Spark.CSharp.Examples.exe Sql.Streaming.StructuredNetworkWordCount localhost 9999

Microsoft.Spark.Examples.Sql.Streaming.StructuredKafkaWordCount (maven accessible)

spark-submit.cmd `
--packages org.apache.spark:spark-sql-kafka-0-10_2.11:2.3.2 `
--class org.apache.spark.deploy.dotnet.DotnetRunner `
--master local `
C:\github\dotnet-spark\src\scala\microsoft-spark-<version>\target\microsoft-spark-<version>.jar `
Microsoft.Spark.CSharp.Examples.exe Sql.Streaming.StructuredKafkaWordCount localhost:9092 subscribe test

Microsoft.Spark.Examples.Sql.Streaming.StructuredKafkaWordCount (jars provided)

spark-submit.cmd 
--jars path\to\net.jpountz.lz4\lz4-1.3.0.jar,path\to\org.apache.kafka\kafka-clients-0.10.0.1.jar,path\to\org.apache.spark\spark-sql-kafka-0-10_2.11-2.3.2.jar,`path\to\org.slf4j\slf4j-api-1.7.6.jar,path\to\org.spark-project.spark\unused-1.0.0.jar,path\to\org.xerial.snappy\snappy-java-1.1.2.6.jar `
--class org.apache.spark.deploy.dotnet.DotnetRunner `
--master local `
C:\github\dotnet-spark\src\scala\microsoft-spark-<version>\target\microsoft-spark-<version>.jar `
Microsoft.Spark.CSharp.Examples.exe Sql.Streaming.StructuredKafkaWordCount localhost:9092 subscribe test

Feel this experience is complicated? Help us by taking up Simplify User Experience for Running an App

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

windows-instructions.md

windows-instructions.md

Building Spark .NET on Windows

Table of Contents

Open Issues:

Pre-requisites:

Building

Building Spark .NET Scala Extensions Layer

Building .NET Samples Application

Using Visual Studio for .NET Framework

Using .NET CLI for .NET 6

Run Samples

Files

windows-instructions.md

Latest commit

History

windows-instructions.md

File metadata and controls

Building Spark .NET on Windows

Table of Contents

Open Issues:

Pre-requisites:

Building

Building Spark .NET Scala Extensions Layer

Building .NET Samples Application

Using Visual Studio for .NET Framework

Using .NET CLI for .NET 6

Run Samples