Pravega

Pravega provides a new storage abstraction - a stream - for continuous and unbounded data. A Pravega stream is an elastic set of durable and append-only segments, each segment being an unbounded sequence of bytes. Streams provide exactly-once semantics, and atomicity for groups of events using transactions.

Project Info: Apache Pekko Connectors Pravega
Artifact	org.apache.pekko pekko-connectors-pravega 1.0.2 Snapshots are available
JDK versions	OpenJDK 8 OpenJDK 11 OpenJDK 17
Scala versions	2.13.14, 2.12.20, 3.3.3
JPMS module name	pekko.stream.connectors.pravega
License	Apache-2.0
API documentation	API (Scaladoc)
Forums	Github Discussions
Release notes	GitHub releases
Issues	Github issues
Sources	https://github.com/apache/pekko-connectors

Artifacts

sbt

libraryDependencies += "org.apache.pekko" %% "pekko-connectors-pravega" % "1.0.2"

Maven

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

Gradle

def versions = [
  ScalaBinary: "2.13"
]
dependencies {
  implementation "org.apache.pekko:pekko-connectors-pravega_${versions.ScalaBinary}:1.0.2"
}

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

Concepts

Pravega stores streams of events, and streams are organized using scopes. A Pravega stream comprises a one or more parallel segments, and the set of parallel segments can change over time with auto-scaling. Pravega is designed to operate at scale and is able to accommodate a large number of segments and streams.

Pravega has an API to write and read events. An application looking into ingesting data writes events to a stream, while consuming data consists of reading events from a stream. In addition to the events API, Pravega has other APIs that enable an application to read and write bytes rather than events and to read events of a stream out of order (e.g., when batch processing).

Pravega stores stream data durably, and applications can access the stream data using the same API both when tailing the stream and when processing past data. The system is architected so that the underlying storage is elastic and it is able to accommodate unbounded streams.

When writing an event, Pravega accepts a routing key parameter, and it guarantees order per key even in the presence of auto-scaling.

Since version 8.x, Pravega supports write and read access to a Key/Value tables.

For more information about Pravega please visit the official documentation.

Configuration

Two categories of properties can/must be provided to configure the connector.

Pravega internals properties that are forwarded to Pravega configuration builders:

• ClientConfig pekko.connectors.pravega.defaults.client-config
• EventWriterConfig pekko.connectors.pravega.writer.config
• ReaderConfig pekko.connectors.pravega.reader.config

Apache Pekko Connectors Connector properties (all others).

reference.conf

copysource# SPDX-License-Identifier: Apache-2.0

pekko.connectors.pravega {
  #
  # ClientConfig (Pravega internals)
  defaults.client-config {
    # ControllerURI The controller rpc URI. This can be of 2 types
    #  1. tcp://ip1:port1,ip2:port2,...
    #   This is used if the controller endpoints are static and can be directly accessed.
    #  2. pravega://ip1:port1,ip2:port2,...
    #   This is used to autodiscovery the controller endpoints from an initial controller list.
    #controller-uri = "tcp://localhost:9090"

    # An optional property representing whether to enable TLS for client's communication with the Controller.
    # If this property and enable-tls-to-segment-store are not set, and the scheme used in controller-uri
    #  is "tls" or "pravegas", TLS is automatically enabled for both client-to-Controller and
    # client-to-Segment Store communications.
    #enable-tls-to-controller = false

    # An optional property representing whether to enable TLS for client's communication with the Controller.
    # If this property and 'enable-tls-to-controller' are not set, and the scheme used in 'controller-uri'
    #   is "tls" or "pravegas", TLS is automatically enabled for both client-to-Controller and
    # client-to-Segment Store communications.
    #enable-tls-to-segment-store = false

    # Maximum number of connections per Segment store to be used by connection pooling.
    #max-connections-per-segment-store=10

    # Path to an optional truststore. If this is null or empty, the default JVM trust store is used.
    # This is currently expected to be a signing certificate for the certification authority.
    #trust-store

    # If the flag 'isEnableTls'  is set, this flag decides whether to enable host name validation or not.
    #validate-host-name=true
  }
  reader {
    client-config = ${pekko.connectors.pravega.defaults.client-config}
    # ReaderConfig (Pravega internals)
    config {
      #disable-time-windows = false
      #initial-allocation-delay = 0
    }
    timeout = 1 second
    # The reader-id must be unique across all instances of a reader group.
    # When a reader-id is not provided one is randomly generated each time a Reader Source is created.
    #reader-id="scala-reader-id"
  }
  writer {
    client-config = ${pekko.connectors.pravega.defaults.client-config}
    maximum-inflight-messages = 10
    # EventWriterConfig (Pravega internals)
    config {
      #automatically-note-time=false
      #backoff-multiple=10
      #enable-connection-pooling=false
      #initial-backoff-millis=1
      #retry-attempts=10
      #
      # The transaction timeout parameter corresponds to the lease renewal period.
      # In every period, the client must send at least one ping to keep the txn alive.
      # If the client fails to do so, then Pravega aborts the txn automatically. The client
      # sends pings internally and requires no application intervention, only that it sets
      # this parameter accordingly.
      #
      # This parameter is additionally used to determine the total amount of time that
      # a txn can remain open. Currently, we set the maximum amount of time for a
      # txn to remain open to be the minimum between 1 day and 1,000 times the value
      # of the lease renewal period. The 1,000 is hardcoded and has been chosen arbitrarily
      # to be a large enough value.
      #
      # The maximum allowed lease time by default is 120s, see:
      #
      # io.pravega.controller.util.Config.PROPERTY_TXN_MAX_LEASE
      #
      # The maximum allowed lease time is a configuration parameter of the controller
      # and can be changed accordingly. Note that being a controller-wide parameter,
      # it affects all transactions.
      #transaction-timeout-time=89999L
    }
  }
  table {
    client-config = ${pekko.connectors.pravega.defaults.client-config}
    maximum-inflight-messages = 10
    # Max entries retrived by iterator.
    max-entries-at-once = 100 
  }

}

The Pravega connector can automatically configure the Pravega client by supplying configuration in an application.conf, or it can be set programmatically with ReaderSettingsBuilderReaderSettingsBuilder or WriterSettingsBuilderWriterSettingsBuilder. See the following sections for examples.

ClientConfig

This configuration holds connection properties (endpoints, protocol) for all communication.

It can be overridden in an application.conf file at the following configuration paths:

• reader: pekko.connectors.pravega.reader.client-config
• writer: pekko.connectors.pravega.writer.client-config

It can be customised programmatically, see below.

Streams

EventReader configuration

A Pravega Source needs a ReaderSettingsReaderSettings to operate, it can be built from configuration and programmatically customized:

Scala

copysourceval readerSettings = ReaderSettingsBuilder(system)
  .clientConfigBuilder(
    _.controllerURI(new URI("pravegas://localhost:9090")) // ClientConfig customization
      .enableTlsToController(true)
      .enableTlsToSegmentStore(true))
  .readerConfigBuilder(_.disableTimeWindows(true)) // ReaderConfig customization
  .withTimeout(3.seconds)
  .withSerializer(new UTF8StringSerializer)

Java

copysourceReaderSettings<String> readerSettings =
    ReaderSettingsBuilder.create(system)
        .clientConfigBuilder(
            builder -> builder.enableTlsToController(true)) // ClientConfig customization
        .readerConfigBuilder(
            builder -> builder.disableTimeWindows(true)) // ReaderConfig customization
        .withTimeout(Duration.ofSeconds(3))
        .withSerializer(new UTF8StringSerializer());

EventWriter configuration

A Pravega Flow or Sink needs a WriterSettingsWriterSettings to operate, it can be built from configuration and programmatically customized:

You may want to use a routing key, you have to provide a key extractor functionkey extractor function for your message type.

Scala

copysourceval writerSettinds = WriterSettingsBuilder(system)
  .clientConfigBuilder(_.enableTlsToController(true)) // ClientConfig customization
  .eventWriterConfigBuilder(_.backoffMultiple(5)) // EventWriterConfig customization
  .withMaximumInflightMessages(5)
  .withKeyExtractor((str: String) => str.substring(0, 2))
  .withSerializer(new UTF8StringSerializer)

Java

copysourceWriterSettings<String> writerSettings =
    WriterSettingsBuilder.<String>create(system)
        .withKeyExtractor((String str) -> str.substring(0, 1))
        .withSerializer(new UTF8StringSerializer());

ReaderSettingsBuilderReaderSettingsBuilder, ReaderSettingsBuilderReaderSettingsBuilder produce respectively ReaderSettings and ReaderSettings once a Serializer is provided.

Writing to Pravega stream

Pravega message writes are done through a Flow/Sink like:

Scala

copysourceSource(1 to 100)
  .map(i => s"event_$i")
  .runWith(Pravega.sink("an_existing_scope", "an_existing_streamName", writerSettings))

Source(1 to 100)
  .map { i =>
    val routingKey = i % 10
    s"${routingKey}_event_$i"
  }
  .runWith(Pravega.sink("an_existing_scope", "an_existing_streamName", writerSettingsWithRoutingKey))

Java

copysourceSink<String, CompletionStage<Done>> sinkWithRouting =
    Pravega.sink("an_existing_scope", "an_existing_scope", writerSettings);

CompletionStage<Done> doneWithRouting =
    Source.from(Arrays.asList("One", "Two", "Three")).runWith(sinkWithRouting, system);

Reading from Pravega stream

First you need to build a reader group:

Scala

copysource
Using(Pravega.readerGroupManager("an_existing_scope", readerSettings.clientConfig)) { readerGroupManager =>
  readerGroupManager.createReaderGroup("myGroup", "stream1", "stream2")
}

Java

copysource
ReaderGroup readerGroup;
try (PravegaReaderGroupManager readerGroupManager =
    Pravega.readerGroup("an_existing_scope", readerSettings.clientConfig())) {
  readerGroup = readerGroupManager.createReaderGroup("my_group", "streamName");
}

Then use this reader group to read from a Source:

Scala

copysource
Pravega
  .source(readerGroup, readerSettings)
  .to(Sink.foreach { (event: PravegaEvent[String]) =>
    val message: String = event.message
    processMessage(message)
  })
  .run()

Java

copysourceCompletionStage<Done> fut =
    Pravega.<String>source(readerGroup, readerSettings)
        .to(Sink.foreach(e -> processMessage(e.message())))
        .run(system);

It produces a stream of PravegaEventPravegaEvent, a thin wrapper which includes some Pravega metadata along with the message.

Key Value Pair table

Since version 0.8 Pravega exposes Key Value Pair table please note that this API is still experimental.

Similarly a Pravega Table Flow or Sink needs a TableWriterSettingsTableWriterSettings to operate:

Scala

copysourceval tableWriterSettings = TableWriterSettingsBuilder[Int, String]()
  .clientConfigBuilder(_.enableTlsToController(true)) // ClientConfig customization
  .withMaximumInflightMessages(5)
  .withKeyExtractor(str => new TableKey(intSerializer.serialize(str.hashCode())))
  .build()

Java

copysourceTableWriterSettings<Integer, String> tableWriterSettings =
    TableWriterSettingsBuilder.<Integer, String>create(
            system, intSerializer, new UTF8StringSerializer())
        .withKeyExtractor(id -> new TableKey(intSerializer.serialize(id)))
        .build();

To read from a Pravega Table Flow or Source needs a TableReaderSettingsTableReaderSettings to operate:

Scala

copysourceval tableReaderSettings = TableReaderSettingsBuilder[Int, String]()
  .clientConfigBuilder(_.enableTlsToController(true)) // ClientConfig customization
  .withMaximumInflightMessages(5)
  .withMaxEntriesAtOnce(100)
  .withKeyExtractor(str => new TableKey(intSerializer.serialize(str.hashCode())))
  .build()

Java

copysourceTableReaderSettings<Integer, String> tableReaderSettings =
    TableReaderSettingsBuilder.<Integer, String>create(
            system, intSerializer, new UTF8StringSerializer())
        .withKeyExtractor(id -> new TableKey(intSerializer.serialize(id)))
        .build();

Writing to Pravega KVP Table

Pravega message writes are done through a Flow/Sink like:

Scala

copysource
// Write through a flow
Source(1 to 10)
  .map(id => (id, Person(id, s"name_$id")))
  .via(PravegaTable.writeFlow("an_existing_scope", "an_existing_tablename", tablewriterSettings))
  .runWith(Sink.ignore)

// Write in a sink
Source(1 to 10)
  .map(id => (id, Person(id, s"name_$id")))
  .runWith(PravegaTable.sink("an_existing_scope", "an_existing_tablename", tablewriterSettings))

Java

copysourcefinal List<Pair<Integer, String>> events =
    Arrays.asList(
        new Pair<Integer, String>(1, "One"),
        new Pair<Integer, String>(2, "Two"),
        new Pair<Integer, String>(3, "Three"),
        new Pair<Integer, String>(4, "Four"));

Sink<Pair<Integer, String>, CompletionStage<Done>> sink =
    PravegaTable.sink("an_existing_scope", "an_existing_tableName", tablewriterSettings);

CompletionStage<Done> done = Source.from(events).toMat(sink, Keep.right()).run(system);

Reading from Pravega KVP Table

Pravega message reads are from a Source:

Scala

copysource
val readingDone = PravegaTable
  .source("an_existing_scope", "an_existing_tablename", tableSettings)
  .to(Sink.foreach(println))
  .run()

Java

copysource
final CompletionStage<Done> pair =
    PravegaTable.source("an_existing_scope", "an_existing_tableName", tableReaderSettings)
        .to(Sink.foreach((TableEntry<String> kvp) -> processKVP(kvp)))
        .run(system);

Or a Flow

Scala

copysourceval tableSettings: TableSettings[String, Int] =
  TableReaderSettingsBuilder[String, Int]()
    .withKeyExtractor(p => new TableKey(stringSerializer.serialize(p)))
    .build()
PravegaTable
  .readFlow[String, Int](
    scope,
    tableName,
    tableSettings)

Support

• Supported channels