Apache Kudu

The Apache Pekko Connectors Kudu connector supports writing to Apache Kudu tables.

Apache Kudu is a free and open source column-oriented data store in the Apache Hadoop ecosystem.

Artifacts

sbt

val PekkoVersion = "1.0.3"
libraryDependencies ++= Seq(
  "org.apache.pekko" %% "pekko-connectors-kudu" % "1.0.2",
  "org.apache.pekko" %% "pekko-stream" % PekkoVersion
)

Maven

<properties>
  <pekko.version>1.0.3</pekko.version>
  <scala.binary.version>2.13</scala.binary.version>
</properties>
<dependencies>
  <dependency>
    <groupId>org.apache.pekko</groupId>
    <artifactId>pekko-connectors-kudu_${scala.binary.version}</artifactId>
    <version>1.0.2</version>
  </dependency>
  <dependency>
    <groupId>org.apache.pekko</groupId>
    <artifactId>pekko-stream_${scala.binary.version}</artifactId>
    <version>${pekko.version}</version>
  </dependency>
</dependencies>

Gradle

def versions = [
  PekkoVersion: "1.0.3",
  ScalaBinary: "2.13"
]
dependencies {
  implementation "org.apache.pekko:pekko-connectors-kudu_${versions.ScalaBinary}:1.0.2"
  implementation "org.apache.pekko:pekko-stream_${versions.ScalaBinary}:${versions.PekkoVersion}"
}

The table below shows direct dependencies of this module and the second tab shows all libraries it depends on transitively.

Configuration

To connect to Kudu you need:

1. Describe the Kudu Schema
2. Define a converter function to map your data type to a PartialRow
3. Specify Kudu CreateTableOptions
4. Set up Apache Pekko Connectors’ KuduTableSettings

Scala

copysource// Kudu Schema
val cols = List(new ColumnSchema.ColumnSchemaBuilder("key", Type.INT32).key(true).build,
  new ColumnSchema.ColumnSchemaBuilder("value", Type.STRING).build)
val schema = new Schema(cols.asJava)

// Converter function
case class Person(id: Int, name: String)
val kuduConverter: Person => PartialRow = { person =>
  val partialRow = schema.newPartialRow()
  partialRow.addInt(0, person.id)
  partialRow.addString(1, person.name)
  partialRow
}

// Kudu table options
val rangeKeys = List("key")
val createTableOptions = new CreateTableOptions().setNumReplicas(1).setRangePartitionColumns(rangeKeys.asJava)

// Pekko Connectors Settings
val kuduTableSettings = KuduTableSettings("test", schema, createTableOptions, kuduConverter)

Java

copysource// Kudu Schema
List<ColumnSchema> columns = new ArrayList<>(2);
columns.add(new ColumnSchema.ColumnSchemaBuilder("key", Type.INT32).key(true).build());
columns.add(new ColumnSchema.ColumnSchemaBuilder("value", Type.STRING).build());
schema = new Schema(columns);

// Converter function
Function<Person, PartialRow> kuduConverter =
    person -> {
      PartialRow partialRow = schema.newPartialRow();
      partialRow.addInt(0, person.id);
      partialRow.addString(1, person.name);
      return partialRow;
    };

// Kudu table options
List<String> rangeKeys = Collections.singletonList("key");
CreateTableOptions createTableOptions =
    new CreateTableOptions().setNumReplicas(1).setRangePartitionColumns(rangeKeys);

// Pekko Connectors Settings
KuduTableSettings<Person> tableSettings =
    KuduTableSettings.create("tablenameSink", schema, createTableOptions, kuduConverter);

The KuduClient by default is automatically managed by the connector. Settings for the client are read from the reference.conf file. A manually initialized client can be injected to the stream using KuduAttributes

Scala

copysourceval masterAddress = "localhost:7051"
val client = new KuduClient.KuduClientBuilder(masterAddress).build
system.registerOnTermination(client.shutdown())

val flow: Flow[Person, Person, NotUsed] =
  KuduTable
    .flow(kuduTableSettings.withTableName("Flow"))
    .withAttributes(KuduAttributes.client(client))

Java

copysourcefinal String masterAddress = "localhost:7051";
final KuduClient client = new KuduClient.KuduClientBuilder(masterAddress).build();
system.registerOnTermination(
    () -> {
      try {
        client.shutdown();
      } catch (KuduException e) {
        e.printStackTrace();
      }
    });

final Flow<Person, Person, NotUsed> flow =
    KuduTable.flow(tableSettings.withTableName("Flow"))
        .withAttributes(KuduAttributes.client(client));

Writing to Kudu in a Flow

Scala

copysourceval flow: Flow[Person, Person, NotUsed] =
  KuduTable.flow(kuduTableSettings.withTableName("Flow"))

val f = Source(11 to 20)
  .map(i => Person(i, s"zozo_$i"))
  .via(flow)
  .runWith(Sink.fold(0)((a, d) => a + d.id))

Java

copysourceFlow<Person, Person, NotUsed> flow = KuduTable.flow(tableSettings.withTableName("Flow"));

CompletionStage<List<Person>> run =
    Source.from(Arrays.asList(200, 201, 202, 203, 204))
        .map((i) -> new Person(i, String.format("name_%d", i)))
        .via(flow)
        .toMat(Sink.seq(), Keep.right())
        .run(system);

Writing to Kudu with a Sink

Scala

copysourceval sink: Sink[Person, Future[Done]] =
  KuduTable.sink(kuduTableSettings.withTableName("Sink"))

val f = Source(1 to 10)
  .map(i => Person(i, s"zozo_$i"))
  .runWith(sink)

Java

copysourcefinal Sink<Person, CompletionStage<Done>> sink =
    KuduTable.sink(tableSettings.withTableName("Sink"));

CompletionStage<Done> o =
    Source.from(Arrays.asList(100, 101, 102, 103, 104))
        .map((i) -> new Person(i, String.format("name %d", i)))
        .runWith(sink, system);