Spark Streaming从Kafka自定义时间间隔内实时统计行数、TopN并将结果存到hbase中

一、统计kafka的topic在10秒间隔内生产数据的行数并将统计结果存入到hbase中

先在hbase中建立相应的表：

create 'linecount','count'

开启kafka集群并建立相应的topic：

[hadoop@h71 kafka_2.10-0.8.2.0]$ bin/kafka-topics.sh --create --zookeeper h71:2181,h72:2181,h73:2181 --replication-factor 3 --partitions 2 --topic test

启动生产者：
[hadoop@h71 kafka_2.10-0.8.2.0]$ bin/kafka-console-producer.sh --broker-list h71:9092,h72:9092,h73:9092 --topic test 

java代码：

import java.text.SimpleDateFormat;
import java.util.Arrays;
import java.util.Date;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Iterator;
import java.util.Map;
import java.util.Set;

import kafka.serializer.StringDecoder;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.hbase.HBaseConfiguration;
import org.apache.hadoop.hbase.client.HTable;
import org.apache.hadoop.hbase.client.Put;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaPairRDD;
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 org.apache.spark.streaming.Durations;
import org.apache.spark.streaming.api.java.JavaDStream;
import org.apache.spark.streaming.api.java.JavaPairDStream;
import org.apache.spark.streaming.api.java.JavaPairInputDStream;
import org.apache.spark.streaming.api.java.JavaStreamingContext;
import org.apache.spark.streaming.kafka.KafkaUtils;

import scala.Tuple2;

public class KafkaDirectWordCountPersistHBase {
private static String beginTime = null;
private static int cishu = 0;
private static int interval = 0;
private static String rowkey = null;
public static Configuration getConfiguration() {
Configuration conf = HBaseConfiguration.create();
conf.set("hbase.rootdir", "hdfs://192.168.8.71:9000/hbase");
conf.set("hbase.zookeeper.quorum", "192.168.8.71");
return conf;
}
public static void insert(String tableName, String rowKey, String family,
String quailifer, String value) {
try {
HTable table = new HTable(getConfiguration(), tableName);
Put put = new Put(rowKey.getBytes());
put.add(family.getBytes(), quailifer.getBytes(), value.getBytes()) ;
table.put(put);
} catch (Exception e) {
e.printStackTrace();
}
}

public static void main(String[] args) {
SparkConf conf = new SparkConf().setAppName("wordcount").setMaster("local[2]");
//这里设置每多少秒计算一次，我这里设置的间隔是10秒
interval = 10;
//		JavaStreamingContext jssc = new JavaStreamingContext(conf, new Duration(10000));	//毫秒
JavaStreamingContext jssc = new JavaStreamingContext(conf,Durations.seconds(interval));	//秒
// 首先要创建一份kafka参数map
Map<String, String> kafkaParams = new HashMap<String, String>();
// 我们这里是不需要zookeeper节点的,所以我们这里放broker.list
kafkaParams.put("metadata.broker.list", "192.168.8.71:9092,192.168.8.72:9092,192.168.8.73:9092");
// 然后创建一个set,里面放入你要读取的Topic,这个就是我们所说的,它给你做的很好,可以并行读取多个topic
Set<String> topics = new HashSet<String>();
topics.add("test");
JavaPairInputDStream<String,String> lines = KafkaUtils.createDirectStream(
jssc,
String.class, // key类型
String.class, // value类型
StringDecoder.class, // 解码器
StringDecoder.class,
kafkaParams,
topics);
//在第一个间隔的时候其实并非一定等于10秒的，而是小于等于10秒的
SimpleDateFormat time = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
java.util.Date date=new java.util.Date();
System.out.println("StreamingContext started->"+time.format(new Date()));
beginTime=time.format(date);

JavaDStream<String> words = lines.flatMap(new FlatMapFunction<Tuple2<String,String>, String>(){
private static final long serialVersionUID = 1L;
@Override
public Iterable<String> call(Tuple2<String,String> tuple) throws Exception {
return Arrays.asList(tuple._2.split("/n"));	//按行进行分隔
}
});

JavaPairDStream<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 {
return new Tuple2<String, Integer>("line", 1);
}
});

JavaPairDStream<String, Integer> wordcounts = pairs.reduceByKey(new Function2<Integer, Integer, Integer>(){
private static final long serialVersionUID = 1L;
@Override
public Integer call(Integer v1, Integer v2) throws Exception {
return v1 + v2;
}
});
wordcounts.print();
wordcounts.foreachRDD(new VoidFunction<JavaPairRDD<String,Integer>>() {
private static final long serialVersionUID = 1L;
@Override
public void call(JavaPairRDD<String, Integer> wordcountsRDD) throws Exception {
SimpleDateFormat time = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
java.util.Date date=new java.util.Date();
System.out.println("endTime1-->"+time.format(new Date()));	//yyyy-MM-dd HH:mm:ss形式
final long endTime1 = System.currentTimeMillis();
System.out.println("endTime1-->"+endTime1);	//时间戳格式
final String endTime=time.format(date);
cishu++;
System.out.println("cishu-->"+cishu);
if(cishu == 1){
rowkey = beginTime+"__"+endTime;
insert("linecount", rowkey, "count", "sum", "0") ;
}else{
SimpleDateFormat hh1 = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
Date date1 = hh1.parse(endTime);
long hb=date1.getTime();
long a2 = hb - interval*1000;
SimpleDateFormat simpleDateFormat = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
Date date2 = new Date(a2);
String beginTime1 = simpleDateFormat.format(date2);
rowkey = beginTime1+"__"+endTime;
insert("linecount", rowkey, "count", "sum", "0") ;
}
//foreachPartition这个方法好像和kafka的topic的分区个数有关系，如果你topic有两个分区，则这个方法会执行两次
wordcountsRDD.foreachPartition(new VoidFunction<Iterator<Tuple2<String,Integer>>>() {
private static final long serialVersionUID = 1L;
@Override
public void call(Iterator<Tuple2<String, Integer>> wordcounts) throws Exception {
Tuple2<String,Integer> wordcount = null;
//注意：这里是利用了在hbase中对同一rowkey同一列再查入数据会覆盖前一次值的特征，所以hbase中linecount表的版本号必须是1，建表的时候如果你不修改版本号的话默认是1
while(wordcounts.hasNext()){
wordcount = wordcounts.next();
insert("linecount", rowkey, "count", "sum", wordcount._2.toString()) ;
}
}
});
}
});
jssc.start();
jssc.awaitTermination();
jssc.close();
}
}

在myeclipse中运行该代码后在kafka的生产者终端输入数据：

hello world

ni hao a

hello spark
注意：如果你是将我这三行复制过去的话还要再按一下回车键，否则的话你实际输入的是两行

过一段时间后再输入数据：

i

love

you

baby

,

come

on

查看linecount表：

hbase(main):187:0> scan 'linecount'
ROW                                                          COLUMN+CELL
2017-07-26 17:27:56__2017-07-26 17:28:00                    column=count:sum, timestamp=1501061244619, value=0
2017-07-26 17:28:00__2017-07-26 17:28:10                    column=count:sum, timestamp=1501061252476, value=3
2017-07-26 17:28:10__2017-07-26 17:28:20                    column=count:sum, timestamp=1501061262405, value=0
2017-07-26 17:28:20__2017-07-26 17:28:30                    column=count:sum, timestamp=1501061272420, value=7
4 row(s) in 0.3150 seconds

二、统计kafka的topic在10秒间隔内生产数据的TopN并将统计结果存入到hbase中

在hbase中创建相应的Top3表：

create 'KafkaTop','TopN'

java代码：

import java.text.SimpleDateFormat;
import java.util.Arrays;
import java.util.Comparator;
import java.util.Date;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Iterator;
import java.util.Map;
import java.util.Set;
import java.util.TreeMap;

import kafka.serializer.StringDecoder;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.hbase.HBaseConfiguration;
import org.apache.hadoop.hbase.client.HTable;
import org.apache.hadoop.hbase.client.Put;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaPairRDD;
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 org.apache.spark.streaming.Durations;
import org.apache.spark.streaming.api.java.JavaDStream;
import org.apache.spark.streaming.api.java.JavaPairDStream;
import org.apache.spark.streaming.api.java.JavaPairInputDStream;
import org.apache.spark.streaming.api.java.JavaStreamingContext;
import org.apache.spark.streaming.kafka.KafkaUtils;

import scala.Tuple2;

/**
* @author huiqiang
* 2017-7-28 11:24
*/
public class KafkaSparkTopN {
private static String beginTime = null;
private static String hbasetable = "KafkaTop";	//将处理结果存到hbase中的表名，在运行程序之前就得存在
private static int cishu = 0;
private static int interval = 10;	//这里设置每多少秒计算一次，我这里设置的间隔是10秒
private static int n = 0;
private static String rowkey = null;
public static int K = 3;	//你想Top几就设置几

//定义treeMap来保持统计结果,由于treeMap是按key升序排列的,这里要人为指定Comparator以实现倒排
public static TreeMap<Integer, String> treeMap = new TreeMap<Integer, String>(new Comparator<Integer>() {
@Override
public int compare(Integer x, Integer y) {
return y.compareTo(x);
}
});

//连接hbase
public static Configuration getConfiguration() {
Configuration conf = HBaseConfiguration.create();
conf.set("hbase.rootdir", "hdfs://192.168.8.71:9000/hbase");
conf.set("hbase.zookeeper.quorum", "192.168.8.71");
return conf;
}
public static void insert2(String tableName,String rowKey,String family,String quailifer,String value){
try {
HTable table1 = new HTable(getConfiguration(), tableName);
Put put = new Put(rowKey.getBytes());
put.add(family.getBytes(), quailifer.getBytes(), value.getBytes());
table1.put(put);
} catch (Exception e) {
e.printStackTrace();
}
}
public static void insert3(String tableName,String rowKey,String family){
try {
HTable table1 = new HTable(getConfiguration(), tableName);
Put put = new Put(rowKey.getBytes());
for (int i = 1; i <= K; i++) {
put.add(family.getBytes(), ("Top"+i).getBytes(), "null".getBytes());
}
table1.put(put);
} catch (Exception e) {
e.printStackTrace();
}
}

public static void main(String[] args) {
SparkConf conf = new SparkConf().setAppName("wordcount").setMaster("local[2]");
//		JavaStreamingContext jssc = new JavaStreamingContext(conf, new Duration(10000));	//毫秒
JavaStreamingContext jssc = new JavaStreamingContext(conf,Durations.seconds(interval));	//秒
// 首先要创建一份kafka参数map
Map<String, String> kafkaParams = new HashMap<String, String>();
// 我们这里是不需要zookeeper节点的,所以我们这里放broker.list
kafkaParams.put("metadata.broker.list", "192.168.8.71:9092,192.168.8.72:9092,192.168.8.73:9092");
// 然后创建一个set,里面放入你要读取的Topic,这个就是我们所说的,它给你做的很好,可以并行读取多个topic
Set<String> topics = new HashSet<String>();
topics.add("test");
JavaPairInputDStream<String,String> lines = KafkaUtils.createDirectStream(
jssc,
String.class, // key类型
String.class, // value类型
StringDecoder.class, // 解码器
StringDecoder.class,
kafkaParams,
topics);
//在第一个间隔的时候其实并非一定等于10秒的，而是小于等于10秒的
SimpleDateFormat time = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
java.util.Date date=new java.util.Date();
System.out.println("StreamingContext started->"+time.format(new Date()));
beginTime=time.format(date);

JavaDStream<String> words = lines.flatMap(new FlatMapFunction<Tuple2<String,String>, String>(){
private static final long serialVersionUID = 1L;
@Override
public Iterable<String> call(Tuple2<String,String> tuple) throws Exception {
return Arrays.asList(tuple._2.split(" "));	//按空格进行分隔
}
});

JavaPairDStream<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 {
return new Tuple2<String, Integer>(word, 1);
}
});

JavaPairDStream<String, Integer> wordcounts = pairs.reduceByKey(new Function2<Integer, Integer, Integer>(){
private static final long serialVersionUID = 1L;
@Override
public Integer call(Integer v1, Integer v2) throws Exception {
return v1 + v2;
}
});
wordcounts.print();
wordcounts.foreachRDD(new VoidFunction<JavaPairRDD<String,Integer>>() {
private static final long serialVersionUID = 1L;
@Override
public void call(JavaPairRDD<String, Integer> wordcountsRDD) throws Exception {
n = 0;
treeMap.clear();
SimpleDateFormat time = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
java.util.Date date=new java.util.Date();
System.out.println("endTime1-->"+time.format(new Date()));	//yyyy-MM-dd HH:mm:ss形式
final long endTime1 = System.currentTimeMillis();
System.out.println("endTime1-->"+endTime1);	//时间戳格式
final String endTime=time.format(date);
cishu++;
System.out.println("cishu-->"+cishu);
if(cishu == 1){
rowkey = beginTime+"__"+endTime;
insert3(hbasetable, rowkey, "TopN");
}else{
SimpleDateFormat hh1 = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
Date date1 = hh1.parse(endTime);
long hb=date1.getTime();
long a2 = hb - interval*1000;
SimpleDateFormat simpleDateFormat = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
Date date2 = new Date(a2);
String beginTime1 = simpleDateFormat.format(date2);
rowkey = beginTime1+"__"+endTime;
insert3(hbasetable, rowkey, "TopN");
}
//foreachPartition这个方法好像和kafka的topic的分区个数有关系，如果你topic有两个分区，则这个方法会执行两次
wordcountsRDD.foreachPartition(new VoidFunction<Iterator<Tuple2<String,Integer>>>() {
private static final long serialVersionUID = 1L;
@Override
public void call(Iterator<Tuple2<String, Integer>> wordcounts) throws Exception {
Tuple2<String,Integer> wordcount = null;
while(wordcounts.hasNext()){
n++;
wordcount = wordcounts.next();
if (treeMap.containsKey(wordcount._2)){
String value = treeMap.get(wordcount._2) + "," + wordcount._1;
treeMap.remove(wordcount._2);
treeMap.put(wordcount._2, value);
}else {
treeMap.put(wordcount._2, wordcount._1);
}
if(treeMap.size() > K) {
treeMap.remove(treeMap.lastKey());
}
}
}
});
if(n!=0){
int y = 0;
for(int num : treeMap.keySet()) {
y++;
//注意：这里是利用了在hbase中对同一rowkey同一列再查入数据会覆盖前一次值的特征，所以hbase中KafkaTop表的版本号必须是1，建表的时候如果你不修改版本号的话默认是1
insert2(hbasetable, rowkey, "TopN", "Top"+y, treeMap.get(num)+" "+num);
}
}
}
});
jssc.start();
jssc.awaitTermination();
jssc.close();
}
}

在myeclipse中运行该代码后在kafka的生产者终端输入数据：

hello world

hello hadoop

hello hive

hello hadoop

hello world

hello world

hbase hive

在myeclipse的打印台会输出：

-------------------------------------------
Time: 1501214340000 ms
-------------------------------------------
(hive,2)
(hello,6)
(world,3)
(hadoop,2)
(hbase,1)
endTime1-->2017-07-28 11:59:00
endTime1-->1501214340455
cishu-->1
。。。。。。省略
-------------------------------------------
Time: 1501214350000 ms
-------------------------------------------
endTime1-->2017-07-28 11:59:10
endTime1-->1501214350090
cishu-->2

查看hbase表：

hbase(main):018:0> scan 'KafkaTop'
ROW                                                          COLUMN+CELL
2017-07-28 11:58:55__2017-07-28 11:59:00                    column=TopN:Top1, timestamp=1501101768643, value=hello 6
2017-07-28 11:58:55__2017-07-28 11:59:00                    column=TopN:Top2, timestamp=1501101768661, value=world 3
2017-07-28 11:58:55__2017-07-28 11:59:00                    column=TopN:Top3, timestamp=1501101768679, value=hadoop,hive 2
2017-07-28 11:59:00__2017-07-28 11:59:10                    column=TopN:Top1, timestamp=1501101770921, value=null
2017-07-28 11:59:00__2017-07-28 11:59:10                    column=TopN:Top2, timestamp=1501101770921, value=null
2017-07-28 11:59:00__2017-07-28 11:59:10                    column=TopN:Top3, timestamp=1501101770921, value=null
2 row(s) in 0.3140 seconds

三、下面这个不是Spark Streaming的，是来自网上的一个列子，相当于离线分析TopN，仅做参考

来自：http://blog.csdn.net/accptanggang/article/details/52924970

下面是源数据hui.txt，我存放在了我的Windows电脑的桌面的spark文件夹里，取出最大的前3个数字：

2

4

1

6

8

10

34

89

java代码：

import java.util.List;
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.Function;
import org.apache.spark.api.java.function.PairFunction;

import scala.Tuple2;

public class SparkTop {
public static void main(String[] args) {
SparkConf conf=new SparkConf().setAppName("Top3").setMaster("local");
JavaSparkContext sc=new JavaSparkContext(conf);

//JavaRDD<String> lines = sc.textFile("hdfs://tgmaster:9000/in/nums2");
JavaRDD<String> lines = sc.textFile("C:\\Users\\huiqiang\\Desktop\\spark\\hui.txt");

//经过map映射，形成键值对的形式。
JavaPairRDD<Integer, Integer> mapToPairRDD = lines.mapToPair(new PairFunction<String, Integer, Integer>() {
private static final long serialVersionUID = 1L;
public Tuple2<Integer, Integer> call(String num) throws Exception {
// TODO Auto-generated method stub
int numObj=Integer.parseInt(num);
Tuple2<Integer, Integer> tuple2 = new Tuple2<Integer, Integer>(numObj, numObj);
return tuple2;
}
});
/**
* 1、通过sortByKey()算子，根据key进行降序排列
* 2、排序完成后，通过map()算子获取排序之后的数字
*/
JavaRDD<Integer> resultRDD = mapToPairRDD.sortByKey(false).map(new Function<Tuple2<Integer,Integer>, Integer>() {
private static final long serialVersionUID = 1L;

public Integer call(Tuple2<Integer, Integer> v1) throws Exception {
// TODO Auto-generated method stub
return v1._1;
}
});
//通过take()算子获取排序后的前3个数字
List<Integer> nums = resultRDD.take(3);
for (Integer num : nums) {
System.out.println(num);
}
sc.close();
}
}

在myeclipse中运行结果为：

89

34

10