Spark 2.0与Cassandra结合 实现Rest API服务

在这篇文章中，我将介绍如何在Spark中使用Akka-http并结合Cassandra实现REST服务，在这个系统中Cassandra用于数据的存储。　　我们已经见识到Spark的威力，如果和Cassandra正确地结合可以实现更强大的系统。我们先创建一个build.sbt文件，内容如下：

name := "cassandra-spark-akka-http-starter-kit"

version := "1.0"

scalaVersion := "2.11.8"

organization := "com.iteblog"

val akkaV = "2.4.5"
libraryDependencies ++ = Seq(
  "org.apache.spark" % "spark-core_2.11" % "2.0.0" ,
  "org.apache.spark" % "spark-sql_2.11" % "2.0.0" ,
  "com.typesafe.akka" %% "akka-http-core" % akkaV,
  "com.typesafe.akka" %% "akka-http-experimental" % akkaV,
  "com.typesafe.akka" %% "akka-http-testkit" % akkaV % "test" ,
  "com.typesafe.akka" %% "akka-http-spray-json-experimental" % akkaV,
  "org.scalatest" %% "scalatest" % "2.2.6" % "test" ,
  "com.datastax.spark" % "spark-cassandra-connector_2.11" % "2.0.0-M3" ,
  "net.liftweb" % "lift-json_2.11" % "2.6.2"

)

assembleArtifact in assemblyPackageScala := false

assemblyMergeStrategy in assembly := {
  case m if m.toLowerCase.endsWith( "manifest.mf" ) = > MergeStrategy.discard
  case m if m.toLowerCase.matches( "meta-inf.*\\.sf$" ) = > MergeStrategy.discard
  case "reference.conf" = > MergeStrategy.concat
  case _ = > MergeStrategy.first
}

ivyScala := ivyScala.value map {
  _ .copy(overrideScalaVersion = true )
}
fork in run := true

上面我们把 assembleArtifact in assemblyPackageScala 设置为false，因为Spark已经包含了Scala library，所以我们不需要再包含了。

样本类User定义

User累仅仅包含id、名字以及Email等信息，定义如下：

package com.iteblog.domain

case class User(id : String, name : String, email : String)

数据访问层

下面代码片段是数据访问层的实现：

package com.iteblog.factories

import com.iteblog.domain.User
import com.typesafe.config.ConfigFactory
import org.apache.spark.SparkConf
import org.apache.spark.sql.SparkSession
import com.datastax.spark.connector. _

import scala.util.Try


trait DatabaseAccess {

  import Context. _

  def create(user : User) : Boolean =
    Try(sc.parallelize(Seq(user)).saveToCassandra(keyspace, tableName)).toOption.isDefined

  def retrieve(id : String) : Option[Array[User]] = Try(sc.cassandraTable[User](keyspace, tableName).where(s "id='$id'" ).collect()).toOption
}

object DatabaseAccess extends DatabaseAccess


object Context {
  val config = ConfigFactory.load()
  val url = config.getString( "cassandra.url" )
  val sparkConf : SparkConf = new SparkConf().setAppName( "Saprk-cassandra-akka-rest-example" ).setMaster( "local[4]" )
    .set( "spark.cassandra.connection.host" , url)
  val spark = SparkSession.builder().config(sparkConf).getOrCreate()
  val sc = spark.sparkContext
  val keyspace = config.getString( "cassandra.keyspace" )
  val tableName = config.getString( "cassandra.tableName" )
}

服务层

下面是路由文件的实现代码：

package com.iteblog.routes

import java.util.UUID

import akka.actor.ActorSystem
import akka.event.Logging
import akka.http.scaladsl.model. _
import akka.http.scaladsl.server.Directives. _
import akka.http.scaladsl.server.{ExceptionHandler, Route}
import akka.stream.ActorMaterializer
import com.iteblog.domain.User
import com.iteblog.factories.DatabaseAccess
import net.liftweb.json. _
import java.util.Date
import net.liftweb.json.Extraction. _


trait SparkService extends DatabaseAccess {

  implicit val system : ActorSystem
  implicit val materializer : ActorMaterializer
  val logger = Logging(system, getClass)


  implicit def myExceptionHandler =
    ExceptionHandler {
      case e : ArithmeticException = >
        extractUri { uri = >
          complete(HttpResponse(StatusCodes.InternalServerError, entity = s "Data is not persisted and something went wrong" ))
        }
    }

  implicit val formats : Formats = new DefaultFormats {
    outer = >
    override val typeHintFieldName = "type"
    override
    val typeHints = ShortTypeHints(List(classOf[String], classOf[Date]))
  }

  val sparkRoutes : Route = {
    get {
      path( "create" / "name" / Segment / "email" / Segment) { (name : String, email : String) = >
        complete {
          val documentId = "user::" + UUID.randomUUID().toString
          try {
            val user = User(documentId,name,email)
            val isPersisted = create(user)
            if (isPersisted) {
              HttpResponse(StatusCodes.Created, entity = s "Data is successfully persisted with id $documentId" )
            } else {
              HttpResponse(StatusCodes.InternalServerError, entity = s "Error found for id : $documentId" )
            }
          } catch {
            case ex : Throwable = >
              logger.error(ex, ex.getMessage)
              HttpResponse(StatusCodes.InternalServerError, entity = s "Error found for id : $documentId" )
          }
        }
      }
    } ~ path( "retrieve" / "id" / Segment) { (listOfIds : String) = >
      get {
        complete {
          try {
            val idAsRDD : Option[Array[User]] = retrieve(listOfIds)
            idAsRDD match {
              case Some(data) = > HttpResponse(StatusCodes.OK, entity = data.headOption.fold( "" )(x = > compact(render(decompose(x)))))
              case None = > HttpResponse(StatusCodes.InternalServerError, entity = s "Data is not fetched and something went wrong" )
            }
          } catch {
            case ex : Throwable = >
              logger.error(ex, ex.getMessage)
              HttpResponse(StatusCodes.InternalServerError, entity = s "Error found for ids : $listOfIds" )
          }
        }
      }
    }
  }
}

服务启动

现在我们需要编写一个用于启动服务的类，其主要目的是启动一个HTTP服务，这样可以供用户调用，如下：

package com.iteblog

import akka.actor.ActorSystem
import akka.http.scaladsl.Http
import akka.stream.ActorMaterializer
import com.iteblog.routes.SparkService
import com.iteblog.factories.Context


class StartSparkServer( implicit val system : ActorSystem,
                       implicit val materializer : ActorMaterializer) extends SparkService {
  def startServer(address : String, port : Int) = {
    Http().bindAndHandle(sparkRoutes, address, port)
  }
}

object StartApplication extends App {
  StartApp
}

object StartApp {
  implicit val system : ActorSystem = ActorSystem( "Spark-Couchbase-Service" )
  implicit val executor = system.dispatcher
  implicit val materializer = ActorMaterializer()
  val server = new StartSparkServer()
  val config = Context.config
  val serverUrl = config.getString( "http.interface" )
  val port = config.getInt( "http.port" )
  server.startServer(serverUrl, port)
}