项目场景：

使用sparkStream接收kafka的数据进行计算，并且打包上传到linux进行spark任务的submit

错误集合：

1.错误1:

Failed to add file:/usr/local/spark-yarn/./myapp/sparkDemo04.jar to Spark environment
java.io.FileNotFoundException: Jar D:\usr\local\spark-yarn\myapp\sparkDemo04.jar not found
WARN ProcfsMetricsGetter: Exception when trying to compute pagesize, as a result reporting of ProcessTree metrics is stopped

2.windows下ideal中在yarn模式下运行代码出错，显示如下报错

WARN CheckpointReader: Error reading checkpoint from file hdfs://hadoop102:9000/checkpoint6/checkpoint-1637834226000
java.io.IOException: java.lang.ClassCastException: cannot assign instance of java.lang.invoke.SerializedLambda to field org.apache.spark.streaming.dstream.MappedDStream.mapFunc of type scala.Function1 in instance of org.apache.spark.streaming.dstream.MappedDStream

3.报的一些kafka包notfound的问题，这个下面就不讨论了，只需要把对应的包下载后放到spark目录下的jars文件中即可，比如常见的

java.lang.NoClassDefFoundError: org/apache/spark/kafka010/KafkaConfigUpdater

都可以通过添加包的方式解决，如果是spark shell里面出现这种错误，则需要在输入spark-shell命令时，在后面添加 --jars 包路径
最初的代码：

import com.study.stream05_kafka.SparkKafka.createSSC
import org.apache.kafka.clients.consumer.{ConsumerConfig, ConsumerRecord}
import org.apache.log4j.{Level, Logger}
import org.apache.spark.SparkConf
import org.apache.spark.rdd.RDD
import org.apache.spark.streaming.dstream.{DStream, InputDStream}
import org.apache.spark.streaming.kafka010.{ConsumerStrategies, KafkaUtils, LocationStrategies}
import org.apache.spark.streaming.{Seconds, StreamingContext}import java.lang.System.getProperty
import scala.collection.mutable.ListBufferobject stream05_kafka {object SparkKafka{def createSSC(): _root_.org.apache.spark.streaming.StreamingContext={//    TODO 创建环境对象//    StreamingContext创建时，第一个参数表示环境配置，第二个是数据采集周期val sparkConf = new SparkConf().setMaster("local[*]").setAppName("kafka2")sparkConf.set("spark.streaming.stopGracefullyOnShutdown","true")sparkConf.set("spark.hadoop.fs.defaultFS","hdfs://hadoop102:9000")sparkConf.set("spark.hadoop.yarn.resoursemanager.address","hadoop103:8088")val streamingContext: StreamingContext = new StreamingContext(sparkConf, Seconds(3))streamingContext.checkpoint("hdfs://hadoop102:9000/checkpoint6")val kafkaPara: Map[String, Object] = Map[String, Object](ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG -> "hadoop102:9092,hadoop103:9092,hadoop104:9092",ConsumerConfig.GROUP_ID_CONFIG -> "second","key.deserializer" -> "org.apache.kafka.common.serialization.StringDeserializer","value.deserializer" -> "org.apache.kafka.common.serialization.StringDeserializer")//    TODO 逻辑处理val kafkaDS: InputDStream[ConsumerRecord[String, String]] = KafkaUtils.createDirectStream[String, String](streamingContext,LocationStrategies.PreferConsistent,ConsumerStrategies.Subscribe[String, String](Set("sparkOnKafka"), kafkaPara))val num: DStream[String] = kafkaDS.map(_.value())val result = num.map(line=>{val flows = line.split(",")val up=flows(1).toIntval down=flows(2).toInt(flows(0),(up,down,up+down))}).updateStateByKey((queueValue, buffValue: Option[(Int,Int,Int)]) => {val cur=buffValue.getOrElse((0,0,0))var curUp=cur._1var curDown=cur._2for (elem <- queueValue) {curUp+=elem._1curDown+=elem._2}Option((curUp,curDown,curUp+curDown))})result.print()streamingContext}}def main(args: Array[String]): Unit = {println("**************")Logger.getLogger("org.apache.spark").setLevel(Level.WARN)System.getProperties.setProperty("HADOOP_USER_NAME", "hadoop")val streamingContext = StreamingContext.getActiveOrCreate("hdfs://hadoop102:9000/checkpoint6", ()=>createSSC())streamingContext.start()//    2.等待关闭streamingContext.awaitTermination()}}

原因分析：

首先，这里指出如果要打包到linux 下在yarn模式下进行spark的submit，需要设置master为yarn，至于是yarn-client还是yarn-cluster需要提交任务时指定，默认是client。我这里写成local，所以一开始都是windows下可以正常连接kafka拿到数据进行计算，但是linux下就不行了。归根结底没有连接yarn。
1.错误1是因为windows下spark任务提交的时候，找不到你的jar包，试想一下spark的spark-submit命令，需要指定jar包以及class
2.这个是序列化问题还是广播变量不适合于检查点的问题，查资料发现广播变量的内容写入hdfs后就难以恢复了，这里可以把错误定位到StreamingContext.getActiveOrCreate里面，这里有时候可以正常进行数据恢复，但是有时候就会报错。解决方法还没找到，我就直接换检查点路径了，一般生产环境下也只有代码升级的情况下会关闭流计算，这里就没有深究，希望大神可以解答一下。猜测是读取检查点数据的时候序列化出了问题

解决方案：

错误1的解决：所以如果要在windows下运行，需要先使用mvn package或者build artifacts对程序进行打包，然后对sparkConf.setJars指定包的路径，这样在windows下就可以正常运行了
错误2的解决：这里我就换检查点了
最后贴一下我最终成功运行的代码

import com.study.stream05_kafka.SparkKafka.createSSC
import org.apache.kafka.clients.consumer.{ConsumerConfig, ConsumerRecord}
import org.apache.log4j.{Level, Logger}
import org.apache.spark.SparkConf
import org.apache.spark.rdd.RDD
import org.apache.spark.streaming.dstream.{DStream, InputDStream}
import org.apache.spark.streaming.kafka010.{ConsumerStrategies, KafkaUtils, LocationStrategies}
import org.apache.spark.streaming.{Seconds, StreamingContext}import java.lang.System.getProperty
import scala.collection.mutable.ListBufferobject stream05_kafka {object SparkKafka{def createSSC(): _root_.org.apache.spark.streaming.StreamingContext={//    TODO 创建环境对象//    StreamingContext创建时，第一个参数表示环境配置，第二个是数据采集周期val sparkConf = new SparkConf().setMaster("yarn").setAppName("kafka2").set("spark.serializer", "org.apache.spark.serializer.KryoSerializer")sparkConf.set("spark.streaming.stopGracefullyOnShutdown","true")sparkConf.set("spark.hadoop.fs.defaultFS","hdfs://hadoop102:9000")sparkConf.set("spark.hadoop.yarn.resoursemanager.address","hadoop103:8088")val streamingContext: StreamingContext = new StreamingContext(sparkConf, Seconds(3))streamingContext.checkpoint("hdfs://hadoop102:9000/checkpoint7")val kafkaPara: Map[String, Object] = Map[String, Object](ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG -> "hadoop102:9092,hadoop103:9092,hadoop104:9092",ConsumerConfig.GROUP_ID_CONFIG -> "second","key.deserializer" -> "org.apache.kafka.common.serialization.StringDeserializer","value.deserializer" -> "org.apache.kafka.common.serialization.StringDeserializer")//    TODO 逻辑处理val kafkaDS: InputDStream[ConsumerRecord[String, String]] = KafkaUtils.createDirectStream[String, String](streamingContext,LocationStrategies.PreferConsistent,ConsumerStrategies.Subscribe[String, String](Set("sparkOnKafka"), kafkaPara))val num: DStream[String] = kafkaDS.map(_.value())val result = num.map(line=>{val flows = line.split(",")val up=flows(1).toIntval down=flows(2).toInt(flows(0),(up,down,up+down))}).updateStateByKey((queueValue, buffValue: Option[(Int,Int,Int)]) => {val cur=buffValue.getOrElse((0,0,0))var curUp=cur._1var curDown=cur._2for (elem <- queueValue) {curUp+=elem._1curDown+=elem._2}Option((curUp,curDown,curUp+curDown))})result.print()streamingContext}}def main(args: Array[String]): Unit = {println("**************")Logger.getLogger("org.apache.spark").setLevel(Level.WARN)System.getProperties.setProperty("HADOOP_USER_NAME", "hadoop")val streamingContext = StreamingContext.getActiveOrCreate("hdfs://hadoop102:9000/checkpoint7", ()=>createSSC())
//    new Thread(new MonitorStop(streamingContext)).start()streamingContext.start()//    2.等待关闭streamingContext.awaitTermination()}}

另外，打包的时候不要添加setJars，否则还是会报错，报的是什么已经忘了，这篇博客也是在我解决问题之后写的，没有记录太多报错，如果我没记错的话可能会报这种错误

cannot assign instance of java.lang.invoke.SerializedLambda to field org.apache.spark.rdd.MapPartitionsRDD.f of type scala.Function3 in instance of org.apache.spark.rdd.MapPartitionsRDD

困惑：

为了解决这个bug，也是在yarn日志和spark日志来回看，看了一天，最让我头疼的就是spark-submit使用control+z退出后，spark-submit进行还会在后台运行，我都怀疑是不是我的kill -9 操作使检查点损坏导致数据恢复失败的，请问各路大神怎么才能结束sparkSubmit进程？

本文来自互联网用户投稿，文章观点仅代表作者本人，不代表本站立场，不承担相关法律责任。如若转载，请注明出处。 如若内容造成侵权/违法违规/事实不符，请点击【内容举报】进行投诉反馈！