正确提交spark到yarn的demo

港控/mmm° 2022-05-24 06:21 252阅读 0赞

通过Spark-submit在xshell提交命令行，如果集群配置了keberos的话需要在打包的jar中进行认证，认证文件上传到节点并且需要分发到每一个节点，节点之间需要无密码ssh登录。

因为是通过Spark-submit提交程序，所以在代码当中的SparkConf设置为

.setMaster("yarn-cluster")

如果提交显示classnotfound可能是当前用户没有权限操作打包在集群上的jar包，或者是缺少命令--master yarn-cluster 。在这里，--master yarn-cluster 要和.setMaster("yarn-cluster")一致，不然会导致节点之间Connection的异常，而xshell一直显示Accepted。

以下是spark-submit提交的命令：

spark-submit  \
     --class sparkDemo.WordCount \
     --master yarn-cluster \
     --num-executors 5 \
     --driver-memory 5g \
     --driver-cores 4 \
     --executor-memory 10g \
     --executor-cores 5 \
      hdfs://1.2.3.4:8020/bulkload/Jars/sub/SparkDemo.jar
    "param1" "param2" "param3"

其中，param是可以传到spark程序main方法的args\[\]。

以下是需要打包的类：

package sparkDemo;
    
    import java.util.Arrays;
    
    import kerberos.KerberosService;
    
    import org.apache.hadoop.conf.Configuration;
    import org.apache.spark.SparkConf;
    import org.apache.spark.api.java.JavaPairRDD;
    import org.apache.spark.api.java.JavaRDD;
    import org.apache.spark.api.java.JavaSparkContext;
    import org.apache.spark.api.java.function.FlatMapFunction;
    import org.apache.spark.api.java.function.Function2;
    import org.apache.spark.api.java.function.PairFunction;
    import org.apache.spark.api.java.function.VoidFunction;
    
    import scala.Tuple2;
    
    import common.HadoopUtil;
    
    public class ParseAttachment {
    	
        public static void main(String[] args) {
        	
    		
    		SparkConf conf_ = new SparkConf()
    				.setMaster("yarn-cluster")
    				.setAppName("parseAttachment");
    
    		JavaSparkContext sc = new JavaSparkContext(conf_);
    		JavaRDD<String> text = sc.textFile(HadoopUtil.hdfs_url + "/bulkload/Spark_in");
    
        	System.out.println("ok");
        	
        	JavaRDD<String> words = text.flatMap(new FlatMapFunction<String, String>() {
                private static final long serialVersionUID = 1L;
                @Override
                public Iterable<String> call(String line) throws Exception {
                    return Arrays.asList(line.split(" "));//把字符串转化成list
                }
            });
            
            JavaPairRDD<String, Integer> pairs = words.mapToPair(new PairFunction<String, String, Integer>() {
                private static final long serialVersionUID = 1L;
                @Override
                public Tuple2<String, Integer> call(String word) throws Exception {
                    // TODO Auto-generated method stub
                    return new Tuple2<String, Integer>(word, 1);
                }
            });
            
            JavaPairRDD<String, Integer> results = pairs.reduceByKey(new Function2<Integer, Integer, Integer>() {            
                private static final long serialVersionUID = 1L;
                @Override
                public Integer call(Integer value1, Integer value2) throws Exception {
                    // TODO Auto-generated method stub
                    return value1 + value2;
                }
            });
            
            JavaPairRDD<Integer, String> temp = results.mapToPair(new PairFunction<Tuple2<String,Integer>, Integer, String>() {
                private static final long serialVersionUID = 1L;
                @Override
                public Tuple2<Integer, String> call(Tuple2<String, Integer> tuple)
                        throws Exception {
                    return new Tuple2<Integer, String>(tuple._2, tuple._1);
                }
            });
            
            JavaPairRDD<String, Integer> sorted = temp.sortByKey(false).mapToPair(new PairFunction<Tuple2<Integer,String>, String, Integer>() {
                private static final long serialVersionUID = 1L;
                @Override
                public Tuple2<String, Integer> call(Tuple2<Integer, String> tuple)
                        throws Exception {
                    // TODO Auto-generated method stub
                    return new Tuple2<String, Integer>(tuple._2,tuple._1);
                }
            });
            
            sorted.foreach(new VoidFunction<Tuple2<String,Integer>>() {
                private static final long serialVersionUID = 1L;
                @Override
                public void call(Tuple2<String, Integer> tuple) throws Exception {
                    System.out.println("word:" + tuple._1 + " count:" + tuple._2);
                }
            });
            
            sc.close();
        }
    }

至于spark的依赖包：[spark-assembly-1.5.2-hadoop2.6.0.jar][] 导进去就可以了。

注：

1.如果报错：the directory item limit is exceed: limit=1048576。因为一般使用spark进行批处理，所以输出的结果文件数量可能非常多，事实上Linux会限制一个文件夹的文件数量，这时候参考：[https://blog.csdn.net/sparkexpert/article/details/51852944][https_blog.csdn.net_sparkexpert_article_details_51852944]修改hdfs-site.xml吧；

2.因为一般处理的文件数量很庞大，所以在代码规范上一定要注意，比如读取文件的流要及时手动关闭，或者通过参数：spark.yarn.executor.memoryOverhead来调节内存分配；

3.报错：[YarnScheduler: Lost executor][YarnScheduler_ Lost executor]。修改执行器等待参数(10分钟)： --conf spark.core.connection.ack.wait.timeout=600。

[spark-assembly-1.5.2-hadoop2.6.0.jar]: https://download.csdn.net/download/huledebing/10189005
[https_blog.csdn.net_sparkexpert_article_details_51852944]: https://blog.csdn.net/sparkexpert/article/details/51852944
[YarnScheduler_ Lost executor]: https://stackoverflow.com/questions/36634829/spark-shell-error-yarnscheduler-lost-executor