我们知道有eclipse的Hadoop插件，能够在eclipse上操作hdfs上的文件和新建mapreduce程序，以及以Run On Hadoop方式运行程序。那么我们可不可以直接在eclipse上运行Spark程序，提交到集群上以YARN-Client方式运行，或者以Standalone方式运行呢？

答案是可以的。下面我来介绍一下如何在eclipse上运行Spark的wordcount程序。我用的hadoop 版本为2.6.2,spark版本为1.5.2。

• 1.Standalone方式运行

• 1.1 新建一个普通的java工程即可，下面直接上代码，

 1 /*   2  * Licensed to the Apache Software Foundation (ASF) under one or more   3  * contributor license agreements.  See the NOTICE file distributed with   4  * this work for additional information regarding copyright ownership.   5  * The ASF licenses this file to You under the Apache License, Version 2.0   6  * (the "License"); you may not use this file except in compliance with   7  * the License.  You may obtain a copy of the License at   8  *   9  *    http://www.apache.org/licenses/LICENSE-2.0  10  *  11  * Unless required by applicable law or agreed to in writing, software  12  * distributed under the License is distributed on an "AS IS" BASIS,  13  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.  14  * See the License for the specific language governing permissions and  15  * limitations under the License.  16  */  17   18 package com.frank.spark;  19   20 import scala.Tuple2;  21 import org.apache.spark.SparkConf;  22 import org.apache.spark.api.java.JavaPairRDD;  23 import org.apache.spark.api.java.JavaRDD;  24 import org.apache.spark.api.java.JavaSparkContext;  25 import org.apache.spark.api.java.function.FlatMapFunction;  26 import org.apache.spark.api.java.function.Function2;  27 import org.apache.spark.api.java.function.PairFunction;  28   29 import java.util.Arrays;  30 import java.util.List;  31 import java.util.regex.Pattern;  32   33 public final class JavaWordCount {  34   private static final Pattern SPACE = Pattern.compile(" ");  35   36   public static void main(String[] args) throws Exception {  37   38     if (args.length < 1) {  39       System.err.println("Usage: JavaWordCount <file>");  40       System.exit(1);  41     }  42   43     SparkConf sparkConf = new SparkConf().setAppName("JavaWordCount");  44     sparkConf.setMaster("spark://192.168.0.1:7077");  45     JavaSparkContext ctx = new JavaSparkContext(sparkConf);  46     ctx.addJar("C:\\Users\\Frank\\sparkwordcount.jar");  47     JavaRDD<String> lines = ctx.textFile(args[0], 1);  48   49     JavaRDD<String> words = lines.flatMap(new FlatMapFunction<String, String>() {  50       @Override  51       public Iterable<String> call(String s) {  52         return Arrays.asList(SPACE.split(s));  53       }  54     });  55   56     JavaPairRDD<String, Integer> ones = words.mapToPair(new PairFunction<String, String, Integer>() {  57       @Override  58       public Tuple2<String, Integer> call(String s) {  59         return new Tuple2<String, Integer>(s, 1);  60       }  61     });  62   63     JavaPairRDD<String, Integer> counts = ones.reduceByKey(new Function2<Integer, Integer, Integer>() {  64       @Override  65       public Integer call(Integer i1, Integer i2) {  66         return i1 + i2;  67       }  68     });  69   70     List<Tuple2<String, Integer>> output = counts.collect();  71     for (Tuple2<?,?> tuple : output) {  72       System.out.println(tuple._1() + ": " + tuple._2());  73     }  74     ctx.stop();  75   }  76 }

代码直接从spark安装包解压后在examples/src/main/java/org/apache/spark/examples/JavaWordCount.java拷贝出来，唯一不同的地方在增加了44行和46行，44行设置了Master,为hadoop的master 结点的IP,端口号为7077。46行设置了工程打包后放置在windows上的路径。

• 1.2 加入spark依赖包spark-assembly-1.5.2-hadoop2.6.0.jar，这个包可以从spark 安装包解压 后在lib目录下。

• 1.3 配置要统计的文件在hdfs上的路径

Run As->Run Configurations

点击Arguments,因为程序中47行要求输入被统计的文件路径，所以在这里配置以下，文件必须放在hdfs上，所以这里的ip也是你的hadoop的master机器的ip.

• 1.4 接下来就是Run程序了，统计的结果会显示在eclipse的控制台。你也可以通过spark的web页面查看刚才提交的程序。

• 2. 以YARN-Client方式运行

• 2.1 先上代码

   1 /*   2  * Licensed to the Apache Software Foundation (ASF) under one or more   3  * contributor license agreements.  See the NOTICE file distributed with   4  * this work for additional information regarding copyright ownership.   5  * The ASF licenses this file to You under the Apache License, Version 2.0   6  * (the "License"); you may not use this file except in compliance with   7  * the License.  You may obtain a copy of the License at   8  *   9  *    http://www.apache.org/licenses/LICENSE-2.0  10  *  11  * Unless required by applicable law or agreed to in writing, software  12  * distributed under the License is distributed on an "AS IS" BASIS,  13  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.  14  * See the License for the specific language governing permissions and  15  * limitations under the License.  16  */  17   18 package com.frank.spark;  19   20 import scala.Tuple2;  21 import org.apache.spark.SparkConf;  22 import org.apache.spark.api.java.JavaPairRDD;  23 import org.apache.spark.api.java.JavaRDD;  24 import org.apache.spark.api.java.JavaSparkContext;  25 import org.apache.spark.api.java.function.FlatMapFunction;  26 import org.apache.spark.api.java.function.Function2;  27 import org.apache.spark.api.java.function.PairFunction;  28   29 import java.util.Arrays;  30 import java.util.List;  31 import java.util.regex.Pattern;  32   33 public final class JavaWordCount {  34   private static final Pattern SPACE = Pattern.compile(" ");  35   36   public static void main(String[] args) throws Exception {  37         38     System.setProperty("HADOOP_USER_NAME", "hadoop");  39   40     if (args.length < 1) {  41       System.err.println("Usage: JavaWordCount <file>");  42       System.exit(1);  43     }  44   45     SparkConf sparkConf = new SparkConf().setAppName("JavaWordCountByFrank01");  46     sparkConf.setMaster("yarn-client");  47     sparkConf.set("spark.yarn.dist.files", "C:\\software\\workspace\\sparkwordcount\\src\\yarn-site.xml");  48     sparkConf.set("spark.yarn.jar", "hdfs://192.168.0.1:9000/user/bigdatagfts/spark-assembly-1.5.2-hadoop2.6.0.jar");  49   50     JavaSparkContext ctx = new JavaSparkContext(sparkConf);  51     ctx.addJar("C:\\Users\\Frank\\sparkwordcount.jar");  52     JavaRDD<String> lines = ctx.textFile(args[0], 1);  53   54     JavaRDD<String> words = lines.flatMap(new FlatMapFunction<String, String>() {  55       @Override  56       public Iterable<String> call(String s) {  57         return Arrays.asList(SPACE.split(s));  58       }  59     });  60   61     JavaPairRDD<String, Integer> ones = words.mapToPair(new PairFunction<String, String, Integer>() {  62       @Override  63       public Tuple2<String, Integer> call(String s) {  64         return new Tuple2<String, Integer>(s, 1);  65       }  66     });  67   68     JavaPairRDD<String, Integer> counts = ones.reduceByKey(new Function2<Integer, Integer, Integer>() {  69       @Override  70       public Integer call(Integer i1, Integer i2) {  71         return i1 + i2;  72       }  73     });  74   75     List<Tuple2<String, Integer>> output = counts.collect();  76     for (Tuple2<?,?> tuple : output) {  77       System.out.println(tuple._1() + ": " + tuple._2());  78     }  79     ctx.stop();  80   }  81 }

• 2.2 程序解释

38行，如果你的windows用户名和集群上用户名不一样，这里就应该配置一下。比如我windows用户名为Frank,而装有hadoop的集群username为hadoop,这里我就以38行这样设置。

46行，这里配置以yarn-client方式

48行，以这种方式运行时候，每一次运行都会把spark-assembly-1.5.2-hadoop2.6.0.jar包上传到hdfs下这次生成的application-id文件夹下，会耗费几分钟时间，这里也可以配置spark.yarn.jar，先把spark-assembly-1.5.2-hadoop2.6.0.jar上传到hdfs一个目录下，这样就不用每次从windows上传到hdfs下了。参考https://spark.apache.org/docs/1.5.2/running-on-yarn.html.

spark.yarn.jar :The location of the Spark jar file, in case overriding the default location is desired. By default, Spark on YARN will use a Spark jar installed locally, but the Spark jar can also be in a world-readable location on HDFS. This allows YARN to cache it on nodes so that it doesn't need to be distributed each time an application runs. To point to a jar on HDFS, for example, set this configuration to "hdfs:///some/path".

51行，把项目打包后放在windows上的路径。

• 2.3 程序配置

把3个配置文件放在src下，配置文件从hadoop的linux机器上拷贝下来。

• 2.4 配置要统计的文件在hdfs上的路径

参考1.3,同样结果显示在eclipse控制台。