flume kafka sparkstreaming整合后集群报错org.apache.kafka.clients.consumer.KafkaConsumer.subscribe(Ljava/uti

简介
     整个项目架构是在CDH中，flume采集数据到kafka，然后sparkstreaming消费（flume1.7版本，kafka0.10版本，spark 2.1版本）。本来local本地模式测试已经没有问题，但是部署到集群上就报错如下：

Exception in thread "streaming-start" java.lang.NoSuchMethodError: org.apache.kafka.clients.consumer.KafkaConsumer.subscribe(Ljava/util/Collection;)V
        at org.apache.spark.streaming.kafka010.Subscribe.onStart(ConsumerStrategy.scala:84)
        at org.apache.spark.streaming.kafka010.DirectKafkaInputDStream.consumer(DirectKafkaInputDStream.scala:75)
        at org.apache.spark.streaming.kafka010.DirectKafkaInputDStream.start(DirectKafkaInputDStream.scala:243)
        at org.apache.spark.streaming.DStreamGraph$$anonfun$start$5.apply(DStreamGraph.scala:49)
        at org.apache.spark.streaming.DStreamGraph$$anonfun$start$5.apply(DStreamGraph.scala:49)
        at scala.collection.parallel.mutable.ParArray$ParArrayIterator.foreach_quick(ParArray.scala:143)
        at scala.collection.parallel.mutable.ParArray$ParArrayIterator.foreach(ParArray.scala:136)
        at scala.collection.parallel.ParIterableLike$Foreach.leaf(ParIterableLike.scala:972)
        at scala.collection.parallel.Task$$anonfun$tryLeaf$1.apply$mcV$sp(Tasks.scala:49)
        at scala.collection.parallel.Task$$anonfun$tryLeaf$1.apply(Tasks.scala:48)
        at scala.collection.parallel.Task$$anonfun$tryLeaf$1.apply(Tasks.scala:48)
        at scala.collection.parallel.Task$class.tryLeaf(Tasks.scala:51)
        at scala.collection.parallel.ParIterableLike$Foreach.tryLeaf(ParIterableLike.scala:969)
        at scala.collection.parallel.AdaptiveWorkStealingTasks$WrappedTask$class.compute(Tasks.scala:152)
        at scala.collection.parallel.AdaptiveWorkStealingForkJoinTasks$WrappedTask.compute(Tasks.scala:443)
        at scala.concurrent.forkjoin.RecursiveAction.exec(RecursiveAction.java:160)
        at scala.concurrent.forkjoin.ForkJoinTask.doExec(ForkJoinTask.java:260)
        at scala.concurrent.forkjoin.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1339)
        at scala.concurrent.forkjoin.ForkJoinPool.runWorker(ForkJoinPool.java:1979)
        at scala.concurrent.forkjoin.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:107)
17/08/15 20:09:30 ERROR yarn.ApplicationMaster: RECEIVED SIGNAL TERM

原因分析
      其实这个在官方文档中有介绍。地址如下：
https://www.cloudera.com/documentation/spark2/latest/topics/spark2_kafka.html#running_jobs
简单说，就是kafka集成spark2，需要在CDH中进行设置。官网介绍了2中方法。这里我采用了第二种，在CDH中进行修改配置的方法。步骤如下：

进入CDH的spark2配置界面，在搜索框中输入SPARK_KAFKA_VERSION，出现如下图，然后选择对应版本，这里我应该选择的是0.10，然后保存配置，重启生效。重新跑sparkstreaming任务，问题解决。