How to 'Chunk and Re-assmble' large messages in Reactive Kafka using Akka-Stream

Question

When sending a large file using Kafka, is it possible to distribute it across partitions and then re-assemble it using Akka-Stream? as described in this presentation:

http://www.slideshare.net/JiangjieQin/handle-large-messages-in-apache-kafka-58692297

Answer 1

The "chunking" side, i.e. the producer, is easy enough to write using something like reactive kafka:

case class LargeMessage(bytes : Seq[Byte], topic : String)

def messageToKafka(message : LargeMessage, maxMessageSize : Int) = 
  Source.fromIterator(() => message.bytes.toIterator)
        .via(Flow[Byte].grouped(maxMessageSize))
        .via(Flow[Seq[Byte]].map(seq => new ProducerRecord(message.topic, seq)))
        .runWith(Producer.plainSink(producerSettings)

The "re-assembling", i.e. the consumer, can be implemented in a manner similar to the documentation:

   val messageFut : Future[LargeMessage] = 
     for {
       bytes <- Consumer.map(_._1).runWith(Sink.seq[Byte])
     } yield LargeMessage(bytes, topic)