在Spark Job Server中保留/共享RDD

Question

我希望持久化来自spark作业的RDD,以便所有后续作业都可以使用Spark Job Server.这是我尝试过的:

工作1:

package spark.jobserver

import com.typesafe.config.{Config, ConfigFactory}
import org.apache.spark._
import org.apache.spark.SparkContext._
import scala.util.Try

object FirstJob extends SparkJob with NamedRddSupport {
  def main(args: Array[String]) {
    val conf = new SparkConf().setMaster("local[4]").setAppName("FirstJob")
    val sc = new SparkContext(conf)
    val config = ConfigFactory.parseString("")
    val results = runJob(sc, config)
    println("Result is " + results)
  }

  override def validate(sc: SparkContext, config: Config): SparkJobValidation = SparkJobValid

  override def runJob(sc: SparkContext, config: Config): Any = {

    // the below variable is to be accessed by other jobs:
    val to_be_persisted : org.apache.spark.rdd.RDD[String] = sc.parallelize(Seq("some text"))

    this.namedRdds.update("resultsRDD", to_be_persisted)
    return to_be_persisted
  }
}

工作2:

package spark.jobserver

import com.typesafe.config.{Config, ConfigFactory}
import org.apache.spark._
import org.apache.spark.SparkContext._
import scala.util.Try


object NextJob extends SparkJob with NamedRddSupport {
  def main(args: Array[String]) {
    val conf = new SparkConf().setMaster("local[4]").setAppName("NextJob")
    val sc = new SparkContext(conf)
    val config = ConfigFactory.parseString("")
    val results = runJob(sc, config)
    println("Result is " + results)
  }

  override def validate(sc: SparkContext, config: Config): SparkJobValidation = SparkJobValid

  override def runJob(sc: SparkContext, config: Config): Any = {

    val rdd = this.namedRdds.get[(String, String)]("resultsRDD").get
    rdd
  }
}

我得到的错误是:

{
  "status": "ERROR",
  "result": {
    "message": "None.get",
    "errorClass": "java.util.NoSuchElementException",
    "stack": ["scala.None$.get(Option.scala:313)", "scala.None$.get(Option.scala:311)", "spark.jobserver.NextJob$.runJob(NextJob.scala:30)", "spark.jobserver.NextJob$.runJob(NextJob.scala:16)", "spark.jobserver.JobManagerActor$$anonfun$spark$jobserver$JobManagerActor$$getJobFuture$4.apply(JobManagerActor.scala:278)", "scala.concurrent.impl.Future$PromiseCompletingRunnable.liftedTree1$1(Future.scala:24)", "scala.concurrent.impl.Future$PromiseCompletingRunnable.run(Future.scala:24)", "java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)", "java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)", "java.lang.Thread.run(Thread.java:745)"]
  }

请修改上面的代码,以便to_be_persisted可以访问.谢谢

编辑:

使用以下方法编译和打包scala源后创建了spark上下文:

curl -d "" 'localhost:8090/contexts/test-context?num-cpu-cores=4&mem-per-node=512m'

使用以下方法调用FirstJob和NextJob:

curl -d "" 'localhost:8090/jobs?appName=test&classPath=spark.jobserver.FirstJob&context=test-context&sync=true'

curl -d "" 'localhost:8090/jobs?appName=test&classPath=spark.jobserver.NextJob&context=test-context&sync=true'

Answer 1

这里似乎有两个问题:

1. 如果您正在使用最新的spark-jobserver版本(0.6.2-SNAPSHOT),那么有一个关于NamedObjects无法正常工作的漏洞 - 似乎符合您的描述:https://github.com/spark-jobserver/spark-jobserver/问题/ 386.

2. 你也有一个小型的不匹配 - 在FirstJob你坚持一个RDD[String],而在NextJob你试图获取一个RDD[(String, String)]- 在NextJob,应该阅读val rdd = this.namedRdds.get[String]("resultsRDD").get).

我已经尝试了你的代码与spark-jobserver版本0.6.0和上述小型修复,它的工作原理.