spark 累加器操作

因为有数据统计的需要，准备使用累加器Accumulator进行统计，totalCount =jssc.sparkContext().accumulator(0, "totalCounter");
然后使用的时候用的是totalCount.add(1),
在local模式的时候是正常的，但是当我用cluster模式的时候会报错:NullPointerException。
这是为什么呢？有碰到过的吗？

版本是否一致，spark环境，jar包，代码中路径等等都核实下。最好贴出来看下

[mw_shl_code=java,true]/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements.  See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License.  You may obtain a copy of the License at
*
*    http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/

package org.apache.spark.examples.streaming;

import java.io.File;
import java.io.IOException;
import java.nio.charset.Charset;
import java.util.Arrays;
import java.util.Iterator;
import java.util.List;
import java.util.regex.Pattern;

import scala.Tuple2;

import com.google.common.io.Files;

import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.api.java.function.*;
import org.apache.spark.broadcast.Broadcast;
import org.apache.spark.streaming.Durations;
import org.apache.spark.streaming.Time;
import org.apache.spark.streaming.api.java.JavaDStream;
import org.apache.spark.streaming.api.java.JavaPairDStream;
import org.apache.spark.streaming.api.java.JavaReceiverInputDStream;
import org.apache.spark.streaming.api.java.JavaStreamingContext;
import org.apache.spark.util.LongAccumulator;

/**
* Use this singleton to get or register a Broadcast variable.
*/
class JavaWordBlacklist {

  private static volatile Broadcast<List<String>> instance = null;

  public static Broadcast<List<String>> getInstance(JavaSparkContext jsc) {
    if (instance == null) {
      synchronized (JavaWordBlacklist.class) {
        if (instance == null) {
          List<String> wordBlacklist = Arrays.asList("a", "b", "c");
          instance = jsc.broadcast(wordBlacklist);
        }
      }
    }
    return instance;
  }
}

/**
* Use this singleton to get or register an Accumulator.
*/
class JavaDroppedWordsCounter {

  private static volatile LongAccumulator instance = null;

  public static LongAccumulator getInstance(JavaSparkContext jsc) {
    if (instance == null) {
      synchronized (JavaDroppedWordsCounter.class) {
        if (instance == null) {
          instance = jsc.sc().longAccumulator("WordsInBlacklistCounter");
        }
      }
    }
    return instance;
  }
}

/**
* Counts words in text encoded with UTF8 received from the network every second. This example also
* shows how to use lazily instantiated singleton instances for Accumulator and Broadcast so that
* they can be registered on driver failures.
*
* Usage: JavaRecoverableNetworkWordCount <hostname> <port> <checkpoint-directory> <output-file>
*   <hostname> and <port> describe the TCP server that Spark Streaming would connect to receive
*   data. <checkpoint-directory> directory to HDFS-compatible file system which checkpoint data
*   <output-file> file to which the word counts will be appended
*
* <checkpoint-directory> and <output-file> must be absolute paths
*
* To run this on your local machine, you need to first run a Netcat server
*
*      `$ nc -lk 9999`
*
* and run the example as
*
*      `$ ./bin/run-example org.apache.spark.examples.streaming.JavaRecoverableNetworkWordCount \
*              localhost 9999 ~/checkpoint/ ~/out`
*
* If the directory ~/checkpoint/ does not exist (e.g. running for the first time), it will create
* a new StreamingContext (will print "Creating new context" to the console). Otherwise, if
* checkpoint data exists in ~/checkpoint/, then it will create StreamingContext from
* the checkpoint data.
*
* Refer to the online documentation for more details.
*/
public final class JavaRecoverableNetworkWordCount {
  private static final Pattern SPACE = Pattern.compile(" ");

  private static JavaStreamingContext createContext(String ip,
                                                    int port,
                                                    String checkpointDirectory,
                                                    String outputPath) {

    // If you do not see this printed, that means the StreamingContext has been loaded
    // from the new checkpoint
    System.out.println("Creating new context");
    final File outputFile = new File(outputPath);
    if (outputFile.exists()) {
      outputFile.delete();
    }
    SparkConf sparkConf = new SparkConf().setAppName("JavaRecoverableNetworkWordCount");
    // Create the context with a 1 second batch size
    JavaStreamingContext ssc = new JavaStreamingContext(sparkConf, Durations.seconds(1));
    ssc.checkpoint(checkpointDirectory);

    // Create a socket stream on target ip:port and count the
    // words in input stream of \n delimited text (eg. generated by 'nc')
    JavaReceiverInputDStream<String> lines = ssc.socketTextStream(ip, port);
    JavaDStream<String> words = lines.flatMap(new FlatMapFunction<String, String>() {
      @Override
      public Iterator<String> call(String x) {
        return Arrays.asList(SPACE.split(x)).iterator();
      }
    });
    JavaPairDStream<String, Integer> wordCounts = words.mapToPair(
      new PairFunction<String, String, Integer>() {
        @Override
        public Tuple2<String, Integer> call(String s) {
          return new Tuple2<>(s, 1);
        }
      }).reduceByKey(new Function2<Integer, Integer, Integer>() {
        @Override
        public Integer call(Integer i1, Integer i2) {
          return i1 + i2;
        }
      });

    wordCounts.foreachRDD(new VoidFunction2<JavaPairRDD<String, Integer>, Time>() {
      @Override
      public void call(JavaPairRDD<String, Integer> rdd, Time time) throws IOException {
        // Get or register the blacklist Broadcast
        final Broadcast<List<String>> blacklist =
            JavaWordBlacklist.getInstance(new JavaSparkContext(rdd.context()));
        // Get or register the droppedWordsCounter Accumulator
        final LongAccumulator droppedWordsCounter =
            JavaDroppedWordsCounter.getInstance(new JavaSparkContext(rdd.context()));
        // Use blacklist to drop words and use droppedWordsCounter to count them
        String counts = rdd.filter(new Function<Tuple2<String, Integer>, Boolean>() {
          @Override
          public Boolean call(Tuple2<String, Integer> wordCount) {
            if (blacklist.value().contains(wordCount._1())) {
              droppedWordsCounter.add(wordCount._2());
              return false;
            } else {
              return true;
            }
          }
        }).collect().toString();
        String output = "Counts at time " + time + " " + counts;
        System.out.println(output);
        System.out.println("Dropped " + droppedWordsCounter.value() + " word(s) totally");
        System.out.println("Appending to " + outputFile.getAbsolutePath());
        Files.append(output + "\n", outputFile, Charset.defaultCharset());
      }
    });

    return ssc;
  }

  public static void main(String[] args) throws Exception {
    if (args.length != 4) {
      System.err.println("You arguments were " + Arrays.asList(args));
      System.err.println(
          "Usage: JavaRecoverableNetworkWordCount <hostname> <port> <checkpoint-directory>\n" +
          "     <output-file>. <hostname> and <port> describe the TCP server that Spark\n" +
          "     Streaming would connect to receive data. <checkpoint-directory> directory to\n" +
          "     HDFS-compatible file system which checkpoint data <output-file> file to which\n" +
          "     the word counts will be appended\n" +
          "\n" +
          "In local mode, <master> should be 'local[n]' with n > 1\n" +
          "Both <checkpoint-directory> and <output-file> must be absolute paths");
      System.exit(1);
    }

    final String ip = args[0];
    final int port = Integer.parseInt(args[1]);
    final String checkpointDirectory = args[2];
    final String outputPath = args[3];

    // Function to create JavaStreamingContext without any output operations
    // (used to detect the new context)
    Function0<JavaStreamingContext> createContextFunc = new Function0<JavaStreamingContext>() {
      @Override
      public JavaStreamingContext call() {
        return createContext(ip, port, checkpointDirectory, outputPath);
      }
    };

    JavaStreamingContext ssc =
      JavaStreamingContext.getOrCreate(checkpointDirectory, createContextFunc);
    ssc.start();
    ssc.awaitTermination();
  }
}[/mw_shl_code]

官方文档中，这样单例模式就可以正常使用，这是为什么呢？还有就是官方默认的Integer可以正常使用，我自定义的String累加器。。就不能正常使用了。。。
private static volatile Accumulator<String>   regionCount = null;
public static Accumulator<String> getRegionCount(JavaSparkContext jsc) {
            if (regionCount == null) {
              synchronized (JavaLineCounter.class) {
                if (regionCount == null) {
                        regionCount = jsc.accumulator("", "regionCounter", new StringAccumulator());
                       
                }
              }
            }
            return regionCount;
          }
这样每次getRegionCount这个方法的时候都是一个新的对象。是空的。