Kafka+Storm+HDFS

cd /myhome/usr/storm

bin/storm nimbus &
bin/storm supervisor &

bin/storm ui &

Kafka+Storm+HDFS整合实践

kafka_2.9.2-0.8.1.1.tgz
apache-storm-0.9.2-incubating.tar.gz

Kafka安装配置

我们使用3台机器搭建Kafka集群：
1
    192.168.4.142   h1
2
    192.168.4.143   h2
3
    192.168.4.144   h3

在安装Kafka集群之前，这里没有使用Kafka自带的Zookeeper，而是独立安装了一个Zookeeper集群，也是使用这3台机器，保证Zookeeper集群正常运行。
首先，在h1上准备Kafka安装文件，执行如下命令：
1
    cd /usr/local/
2
    wget http://mirror.bit.edu.cn/apache/kafka/0.8.1.1/kafka_2.9.2-0.8.1.1.tgz
3
    tar xvzf kafka_2.9.2-0.8.1.1.tgz
4
    ln -s /usr/local/kafka_2.9.2-0.8.1.1 /usr/local/kafka
5
    chown -R kafka:kafka /usr/local/kafka_2.9.2-0.8.1.1 /usr/local/kafka

修改配置文件/usr/local/kafka/config/server.properties，修改如下内容：
1
    broker.id=0
2
    zookeeper.connect=h1:2181,h2:2181,h3:2181/kafka

这里需要说明的是，默认Kafka会使用ZooKeeper默认的/路径，这样有关Kafka的ZooKeeper配置就会散落在根路径下面，如果你有其他的应用也在使用ZooKeeper集群，查看ZooKeeper中数据可能会不直观，所以强烈建议指定一个chroot路径，直接在zookeeper.connect配置项中指定：
1
    zookeeper.connect=h1:2181,h2:2181,h3:2181/kafka

而且，需要手动在ZooKeeper中创建路径/kafka，使用如下命令连接到任意一台ZooKeeper服务器：
1
    cd /usr/local/zookeeper
2
    bin/zkCli.sh

在ZooKeeper执行如下命令创建chroot路径：
1
    create /kafka ''

这样，每次连接Kafka集群的时候（使用--zookeeper选项），也必须使用带chroot路径的连接字符串，后面会看到。
然后，将配置好的安装文件同步到其他的h2、h3节点上：
1
    scp -r /usr/local/kafka_2.9.2-0.8.1.1/ h2:/usr/local/
2
    scp -r /usr/local/kafka_2.9.2-0.8.1.1/ h3:/usr/local/

最后，在h2、h3节点上配置，执行如下命令：
1
    cd /usr/local/

kafka/bin/kafka-topics.sh --list --zookeeper 192.168.3.130:2181 192.168.3.140:2181 192.168.3.142:2181

kafka/bin/kafka-topics.sh --describe --zookeeper 192.168.3.130:2181 192.168.3.140:2181 192.168.3.142:2181

2
    ln -s /usr/local/kafka_2.9.2-0.8.1.1 /usr/local/kafka
3
    chown -R kafka:kafka /usr/local/kafka_2.9.2-0.8.1.1 /usr/local/kafka

并修改配置文件/usr/local/kafka/config/server.properties内容如下所示：
1
    broker.id=1  # 在h1修改
2
    
3
    broker.id=2  # 在h2修改

因为Kafka集群需要保证各个Broker的id在整个集群中必须唯一，需要调整这个配置项的值（如果在单机上，可以通过建立多个Broker进程来模拟分布式的Kafka集群，也需要Broker的id唯一，还需要修改一些配置目录的信息）。
在集群中的h1、h2、h3这三个节点上分别启动Kafka，分别执行如下命令：
1
    bin/kafka-server-start.sh /usr/local/kafka/config/server.properties &

可以通过查看日志，或者检查进程状态，保证Kafka集群启动成功。
我们创建一个名称为my-replicated-topic5的Topic，5个分区，并且复制因子为3，执行如下命令：
1
    bin/kafka-topics.sh --create --zookeeper h1:2181,h2:2181,h3:2181/kafka --replication-factor 3 --partitions 5 --topic my-replicated-topic5

查看创建的Topic，执行如下命令：
1
    bin/kafka-topics.sh --describe --zookeeper h1:2181,h2:2181,h3:2181/kafka --topic my-replicated-topic5

结果信息如下所示：
1
    Topic:my-replicated-topic5     PartitionCount:5     ReplicationFactor:3     Configs:
2
         Topic: my-replicated-topic5     Partition: 0     Leader: 0     Replicas: 0,2,1     Isr: 0,2,1
3
         Topic: my-replicated-topic5     Partition: 1     Leader: 0     Replicas: 1,0,2     Isr: 0,2,1
4
         Topic: my-replicated-topic5     Partition: 2     Leader: 2     Replicas: 2,1,0     Isr: 2,0,1
5
         Topic: my-replicated-topic5     Partition: 3     Leader: 0     Replicas: 0,1,2     Isr: 0,2,1
6
         Topic: my-replicated-topic5     Partition: 4     Leader: 2     Replicas: 1,2,0     Isr: 2,0,1

上面Leader、Replicas、Isr的含义如下：
1
    Partition： 分区
2
    Leader   ： 负责读写指定分区的节点
3
    Replicas ： 复制该分区log的节点列表
4
    Isr      ： "in-sync" replicas，当前活跃的副本列表（是一个子集），并且可能成为Leader

我们可以通过Kafka自带的bin/kafka-console-producer.sh和bin/kafka-console-consumer.sh脚本，来验证演示如果发布消息、消费消息。
在一个终端，启动Producer，并向我们上面创建的名称为my-replicated-topic5的Topic中生产消息，执行如下脚本：
1
    bin/kafka-console-producer.sh --broker-list h1:9092,h2:9092,h3:9092 --topic my-replicated-topic5

在另一个终端，启动Consumer，并订阅我们上面创建的名称为my-replicated-topic5的Topic中生产的消息，执行如下脚本：
1
    bin/kafka-console-consumer.sh --zookeeper h1:2181,h2:2181,h3:2181/kafka --from-beginning --topic my-replicated-topic5

可以在Producer终端上输入字符串消息行，然后回车，就可以在Consumer终端上看到消费者消费的消息内容。
也可以参考Kafka的Producer和Consumer的Java API，通过API编码的方式来实现消息生产和消费的处理逻辑。

Storm安装配置

Storm集群也依赖Zookeeper集群，要保证Zookeeper集群正常运行。Storm的安装配置比较简单，我们仍然使用下面3台机器搭建：
1
    192.168.4.142   h1
2
    192.168.4.143   h2
3
    192.168.4.144   h3

首先，在h1节点上，执行如下命令安装：
1
    cd /usr/local/
2
    wget http://mirror.bit.edu.cn/apache/incubator/storm/apache-storm-0.9.2-incubating/apache-storm-0.9.2-incubating.tar.gz
3
    tar xvzf apache-storm-0.9.2-incubating.tar.gz
4
    ln -s /usr/local/apache-storm-0.9.2-incubating /usr/local/storm
5
    chown -R storm:storm /usr/local/apache-storm-0.9.2-incubating /usr/local/storm

然后，修改配置文件conf/storm.yaml，内容如下所示：
01
    storm.zookeeper.servers:
02
         - "h1"
03
         - "h2"
04
         - "h3"
05
    storm.zookeeper.port: 2181
06
    #
07
    nimbus.host: "h1"
08
    
09
    supervisor.slots.ports:
10
        - 6700
11
        - 6701
12
        - 6702
13
        - 6703
14
    
15
    storm.local.dir: "/tmp/storm"

将配置好的安装文件，分发到其他节点上：
1
    scp -r /usr/local/apache-storm-0.9.2-incubating/ h2:/usr/local/
2
    scp -r /usr/local/apache-storm-0.9.2-incubating/ h3:/usr/local/

最后，在h2、h3节点上配置，执行如下命令：
1
    cd /usr/local/
2
    ln -s /usr/local/apache-storm-0.9.2-incubating /usr/local/storm
3
    chown -R storm:storm /usr/local/apache-storm-0.9.2-incubating /usr/local/storm

Storm集群的主节点为Nimbus，从节点为Supervisor，我们需要在h1上启动Nimbus服务，在从节点h2、h3上启动Supervisor服务：
1
    bin/storm nimbus &
2
    bin/storm supervisor &

为了方便监控，可以启动Storm UI，可以从Web页面上监控Storm Topology的运行状态，例如在h2上启动：
1
    bin/storm ui &

这样可以通过访问http://h2:8080/来查看Topology的运行状况。

整合Kafka+Storm

消息通过各种方式进入到Kafka消息中间件，比如可以通过使用Flume来收集日志数据，然后在Kafka中路由暂存，
然后再由实时计算程序Storm做实时分析，这时我们就需要将在Storm的Spout中读取Kafka中的消息，然后交由具体的Spot组件去分析处理。
实际上，apache-storm-0.9.2-incubating这个版本的Storm已经自带了一个集成Kafka的外部插件程序storm-kafka，可以直接使用，
例如我使用的Maven依赖配置，如下所示：
org.apache.stormstorm-core0.9.2-incubatingprovidedorg.apache.stormstorm-kafka0.9.2-incubatingorg.apache.kafkakafka_2.9.20.8.1.1org.apache.zookeeperzookeeperlog4jlog4j

下面，我们开发了一个简单WordCount示例程序，从Kafka读取订阅的消息行，通过空格拆分出单个单词，
然后再做词频统计计算，实现的Topology的代码，如下所示：

package org.shirdrn.storm.examples;

import java.util.Arrays;
import java.util.HashMap;
import java.util.Iterator;
import java.util.Map;
import java.util.Map.Entry;
import java.util.concurrent.atomic.AtomicInteger;

import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;

import storm.kafka.BrokerHosts;
import storm.kafka.KafkaSpout;
import storm.kafka.SpoutConfig;
import storm.kafka.StringScheme;
import storm.kafka.ZkHosts;
import backtype.storm.Config;
import backtype.storm.LocalCluster;
import backtype.storm.StormSubmitter;
import backtype.storm.generated.AlreadyAliveException;
import backtype.storm.generated.InvalidTopologyException;
import backtype.storm.spout.SchemeAsMultiScheme;
import backtype.storm.task.OutputCollector;
import backtype.storm.task.TopologyContext;
import backtype.storm.topology.OutputFieldsDeclarer;
import backtype.storm.topology.TopologyBuilder;
import backtype.storm.topology.base.BaseRichBolt;
import backtype.storm.tuple.Fields;
import backtype.storm.tuple.Tuple;
import backtype.storm.tuple.Values;

public class MyKafkaTopology {

     public static class KafkaWordSplitter extends BaseRichBolt {

          private static final Log LOG = LogFactory.getLog(KafkaWordSplitter.class);
          private static final long serialVersionUID = 886149197481637894L;
          private OutputCollector collector;
        
          @Override
          public void prepare(Map stormConf, TopologyContext context,
                    OutputCollector collector) {
               this.collector = collector;             
          }

          @Override
          public void execute(Tuple input) {
               String line = input.getString(0);
               LOG.info("RECV[kafka -> splitter] " + line);
               String[] words = line.split("\\s+");
               for(String word : words) {
                    LOG.info("EMIT[splitter -> counter] " + word);
                    collector.emit(input, new Values(word, 1));
               }
               collector.ack(input);
          }

          @Override
          public void declareOutputFields(OutputFieldsDeclarer declarer) {
               declarer.declare(new Fields("word", "count"));        
          }
        
     }
   
     public static class WordCounter extends BaseRichBolt {

          private static final Log LOG = LogFactory.getLog(WordCounter.class);
          private static final long serialVersionUID = 886149197481637894L;
          private OutputCollector collector;
          private Map counterMap;
        
          @Override
          public void prepare(Map stormConf, TopologyContext context,
                    OutputCollector collector) {
               this.collector = collector;   
               this.counterMap = new HashMap();
          }

          @Override
          public void execute(Tuple input) {
               String word = input.getString(0);
               int count = input.getInteger(1);
               LOG.info("RECV[splitter -> counter] " + word + " : " + count);
               AtomicInteger ai = this.counterMap.get(word);
               if(ai == null) {
                    ai = new AtomicInteger();
                    this.counterMap.put(word, ai);
               }
               ai.addAndGet(count);
               collector.ack(input);
               LOG.info("CHECK statistics map: " + this.counterMap);
          }

          @Override
          public void cleanup() {
               LOG.info("The final result:");
               Iterator> iter = this.counterMap.entrySet().iterator();
               while(iter.hasNext()) {
                    Entry entry = iter.next();
                    LOG.info(entry.getKey() + "\t:\t" + entry.getValue().get());
               }
             
          }

          @Override
          public void declareOutputFields(OutputFieldsDeclarer declarer) {
               declarer.declare(new Fields("word", "count"));        
          }
     }
   
     public static void main(String[] args) throws AlreadyAliveException, InvalidTopologyException, InterruptedException {
          String zks = "h1:2181,h2:2181,h3:2181";
          String topic = "my-replicated-topic5";
          String zkRoot = "/storm"; // default zookeeper root configuration for storm
          String id = "word";
        
          BrokerHosts brokerHosts = new ZkHosts(zks);
          SpoutConfig spoutConf = new SpoutConfig(brokerHosts, topic, zkRoot, id);
          spoutConf.scheme = new SchemeAsMultiScheme(new StringScheme());
          spoutConf.forceFromStart = false;
          spoutConf.zkServers = Arrays.asList(new String[] {"h1", "h2", "h3"});
          spoutConf.zkPort = 2181;
        
          TopologyBuilder builder = new TopologyBuilder();
          builder.setSpout("kafka-reader", new KafkaSpout(spoutConf), 5); // Kafka我们创建了一个5分区的Topic，这里并行度设置为5
          builder.setBolt("word-splitter", new KafkaWordSplitter(), 2).shuffleGrouping("kafka-reader");
          builder.setBolt("word-counter", new WordCounter()).fieldsGrouping("word-splitter", new Fields("word"));
        
          Config conf = new Config();
        
          String name = MyKafkaTopology.class.getSimpleName();
          if (args != null && args.length > 0) {
               // Nimbus host name passed from command line
               conf.put(Config.NIMBUS_HOST, args[0]);
               conf.setNumWorkers(3);
               StormSubmitter.submitTopologyWithProgressBar(name, conf, builder.createTopology());
          } else {
               conf.setMaxTaskParallelism(3);
               LocalCluster cluster = new LocalCluster();
               cluster.submitTopology(name, conf, builder.createTopology());
               Thread.sleep(60000);
               cluster.shutdown();
          }
     }
}


上面程序，在本地调试（使用LocalCluster）不需要输入任何参数，提交到实际集群中运行时，需要传递一个参数，该参数为Nimbus的主机名称。
通过Maven构建，生成一个包含依赖的single jar文件（不要把Storm的依赖包添加进去），例如storm-examples-0.0.1-SNAPSHOT.jar，
在提交Topology程序到Storm集群之前，因为用到了Kafka，需要拷贝一下依赖jar文件到Storm集群中的lib目录下面：

cp /usr/local/kafka/libs/kafka_2.9.2-0.8.1.1.jar /usr/local/storm/lib/
cp /usr/local/kafka/libs/scala-library-2.9.2.jar /usr/local/storm/lib/
cp /usr/local/kafka/libs/metrics-core-2.2.0.jar /usr/local/storm/lib/
cp /usr/local/kafka/libs/snappy-java-1.0.5.jar /usr/local/storm/lib/
cp /usr/local/kafka/libs/zkclient-0.3.jar /usr/local/storm/lib/
cp /usr/local/kafka/libs/log4j-1.2.15.jar /usr/local/storm/lib/
cp /usr/local/kafka/libs/slf4j-api-1.7.2.jar /usr/local/storm/lib/
cp /usr/local/kafka/libs/jopt-simple-3.2.jar /usr/local/storm/lib/
然后，就可以提交我们开发的Topology程序了：
bin/storm jar /home/storm/storm-examples-0.0.1-SNAPSHOT.jar org.shirdrn.storm.examples.MyKafkaTopology h1

可以通过查看日志文件（logs/目录下）或者Storm UI来监控Topology的运行状况。如果程序没有错误，可以使用前面我们使用的Kafka Producer来生成消息，就能看到我们开发的Storm Topology能够实时接收到并进行处理。
上面Topology实现代码中，有一个很关键的配置对象SpoutConfig，配置属性如下所示：
spoutConf.forceFromStart = false;
该配置是指，如果该Topology因故障停止处理，下次正常运行时是否从Spout对应数据源Kafka中的该订阅Topic的起始位置开始读取，如果forceFromStart=true，则之前处理过的Tuple还要重新处理一遍，否则会从上次处理的位置继续处理，保证Kafka中的Topic数据不被重复处理，是在数据源的位置进行状态记录。

整合Storm+HDFS

Storm实时计算集群从Kafka消息中间件中消费消息，有实时处理需求的可以走实时处理程序，还有需要进行离线分析的需求，如写入到HDFS进行分析。下面实现了一个Topology，代码如下所示：

package org.shirdrn.storm.examples;

import java.text.DateFormat;
import java.text.SimpleDateFormat;
import java.util.Date;
import java.util.Map;
import java.util.Random;

import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.storm.hdfs.bolt.HdfsBolt;
import org.apache.storm.hdfs.bolt.format.DefaultFileNameFormat;
import org.apache.storm.hdfs.bolt.format.DelimitedRecordFormat;
import org.apache.storm.hdfs.bolt.format.FileNameFormat;
import org.apache.storm.hdfs.bolt.format.RecordFormat;
import org.apache.storm.hdfs.bolt.rotation.FileRotationPolicy;
import org.apache.storm.hdfs.bolt.rotation.TimedRotationPolicy;
import org.apache.storm.hdfs.bolt.rotation.TimedRotationPolicy.TimeUnit;
import org.apache.storm.hdfs.bolt.sync.CountSyncPolicy;
import org.apache.storm.hdfs.bolt.sync.SyncPolicy;

import backtype.storm.Config;
import backtype.storm.LocalCluster;
import backtype.storm.StormSubmitter;
import backtype.storm.generated.AlreadyAliveException;
import backtype.storm.generated.InvalidTopologyException;
import backtype.storm.spout.SpoutOutputCollector;
import backtype.storm.task.TopologyContext;
import backtype.storm.topology.OutputFieldsDeclarer;
import backtype.storm.topology.TopologyBuilder;
import backtype.storm.topology.base.BaseRichSpout;
import backtype.storm.tuple.Fields;
import backtype.storm.tuple.Values;
import backtype.storm.utils.Utils;

public class StormToHDFSTopology {

     public static class EventSpout extends BaseRichSpout {

          private static final Log LOG = LogFactory.getLog(EventSpout.class);
          private static final long serialVersionUID = 886149197481637894L;
          private SpoutOutputCollector collector;
          private Random rand;
          private String[] records;
        
          @Override
          public void open(Map conf, TopologyContext context,
                    SpoutOutputCollector collector) {
               this.collector = collector;   
               rand = new Random();
               records = new String[] {
                         "10001     ef2da82d4c8b49c44199655dc14f39f6     4.2.1     HUAWEI G610-U00     HUAWEI     2     70:72:3c:73:8b:22     2014-10-13 12:36:35",
                         "10001     ffb52739a29348a67952e47c12da54ef     4.3     GT-I9300     samsung     2     50:CC:F8:E4:22:E2     2014-10-13 12:36:02",
                         "10001     ef2da82d4c8b49c44199655dc14f39f6     4.2.1     HUAWEI G610-U00     HUAWEI     2     70:72:3c:73:8b:22     2014-10-13 12:36:35"
               };
          }


          @Override
          public void nextTuple() {
               Utils.sleep(1000);
               DateFormat df = new SimpleDateFormat("yyyy-MM-dd_HH-mm-ss");
               Date d = new Date(System.currentTimeMillis());
               String minute = df.format(d);
               String record = records[rand.nextInt(records.length)];
               LOG.info("EMIT[spout -> hdfs] " + minute + " : " + record);
               collector.emit(new Values(minute, record));
          }

          @Override
          public void declareOutputFields(OutputFieldsDeclarer declarer) {
               declarer.declare(new Fields("minute", "record"));        
          }


     }
   
     public static void main(String[] args) throws AlreadyAliveException, InvalidTopologyException, InterruptedException {
          // use "|" instead of "," for field delimiter
          RecordFormat format = new DelimitedRecordFormat()
                  .withFieldDelimiter(" : ");

          // sync the filesystem after every 1k tuples
          SyncPolicy syncPolicy = new CountSyncPolicy(1000);

          // rotate files
          FileRotationPolicy rotationPolicy = new TimedRotationPolicy(1.0f, TimeUnit.MINUTES);

          FileNameFormat fileNameFormat = new DefaultFileNameFormat()
                  .withPath("/storm/").withPrefix("app_").withExtension(".log");

          HdfsBolt hdfsBolt = new HdfsBolt()
                  .withFsUrl("hdfs://h1:8020")
                  .withFileNameFormat(fileNameFormat)
                  .withRecordFormat(format)
                  .withRotationPolicy(rotationPolicy)
                  .withSyncPolicy(syncPolicy);
        
          TopologyBuilder builder = new TopologyBuilder();
          builder.setSpout("event-spout", new EventSpout(), 3);
          builder.setBolt("hdfs-bolt", hdfsBolt, 2).fieldsGrouping("event-spout", new Fields("minute"));
        
          Config conf = new Config();
        
          String name = StormToHDFSTopology.class.getSimpleName();
          if (args != null && args.length > 0) {
               conf.put(Config.NIMBUS_HOST, args[0]);
               conf.setNumWorkers(3);
               StormSubmitter.submitTopologyWithProgressBar(name, conf, builder.createTopology());
          } else {
               conf.setMaxTaskParallelism(3);
               LocalCluster cluster = new LocalCluster();
               cluster.submitTopology(name, conf, builder.createTopology());
               Thread.sleep(60000);
               cluster.shutdown();
          }
     }

}
上面的处理逻辑，可以对HdfsBolt进行更加详细的配置，如FileNameFormat、SyncPolicy、FileRotationPolicy（可以设置在满足什么条件下，切出一个新的日志，如可以指定多长时间切出一个新的日志文件，可以指定一个日志文件大小达到设置值后，再写一个新日志文件），更多设置可以参考storm-hdfs，。
上面代码在打包的时候，需要注意，使用storm-starter自带的Maven打包配置，可能在将Topology部署运行的时候，会报错，可以使用maven-shade-plugin这个插件，如下配置所示：
org.apache.maven.pluginsmaven-shade-plugin1.4truepackageshade
整合Kafka+Storm+HDFS

上面分别对整合Kafka+Storm和Storm+HDFS做了实践，可以将后者的Spout改成前者的Spout，从Kafka中消费消息，在Storm中可以做简单处理，
然后将数据写入HDFS，最后可以在Hadoop平台上对数据进行离线分析处理。下面，写了一个简单的例子，从Kafka消费消息，然后经由Storm处理，
写入到HDFS存储，代码如下所示：
package org.shirdrn.storm.examples;

import java.util.Arrays;
import java.util.Map;

import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.storm.hdfs.bolt.HdfsBolt;
import org.apache.storm.hdfs.bolt.format.DefaultFileNameFormat;
import org.apache.storm.hdfs.bolt.format.DelimitedRecordFormat;
import org.apache.storm.hdfs.bolt.format.FileNameFormat;
import org.apache.storm.hdfs.bolt.format.RecordFormat;
import org.apache.storm.hdfs.bolt.rotation.FileRotationPolicy;
import org.apache.storm.hdfs.bolt.rotation.TimedRotationPolicy;
import org.apache.storm.hdfs.bolt.rotation.TimedRotationPolicy.TimeUnit;
import org.apache.storm.hdfs.bolt.sync.CountSyncPolicy;
import org.apache.storm.hdfs.bolt.sync.SyncPolicy;

import storm.kafka.BrokerHosts;
import storm.kafka.KafkaSpout;
import storm.kafka.SpoutConfig;
import storm.kafka.StringScheme;
import storm.kafka.ZkHosts;
import backtype.storm.Config;
import backtype.storm.LocalCluster;
import backtype.storm.StormSubmitter;
import backtype.storm.generated.AlreadyAliveException;
import backtype.storm.generated.InvalidTopologyException;
import backtype.storm.spout.SchemeAsMultiScheme;
import backtype.storm.task.OutputCollector;
import backtype.storm.task.TopologyContext;
import backtype.storm.topology.OutputFieldsDeclarer;
import backtype.storm.topology.TopologyBuilder;
import backtype.storm.topology.base.BaseRichBolt;
import backtype.storm.tuple.Fields;
import backtype.storm.tuple.Tuple;
import backtype.storm.tuple.Values;

public class DistributeWordTopology {
   
     public static class KafkaWordToUpperCase extends BaseRichBolt {

          private static final Log LOG = LogFactory.getLog(KafkaWordToUpperCase.class);
          private static final long serialVersionUID = -5207232012035109026L;
          private OutputCollector collector;
        
          @Override
          public void prepare(Map stormConf, TopologyContext context,
                    OutputCollector collector) {
               this.collector = collector;             
          }

          @Override
          public void execute(Tuple input) {
               String line = input.getString(0).trim();
               LOG.info("RECV[kafka -> splitter] " + line);
               if(!line.isEmpty()) {
                    String upperLine = line.toUpperCase();
                    LOG.info("EMIT[splitter -> counter] " + upperLine);
                    collector.emit(input, new Values(upperLine, upperLine.length()));
               }
               collector.ack(input);
          }

          @Override
          public void declareOutputFields(OutputFieldsDeclarer declarer) {
               declarer.declare(new Fields("line", "len"));        
          }
        
     }
   
     public static class RealtimeBolt extends BaseRichBolt {

          private static final Log LOG = LogFactory.getLog(KafkaWordToUpperCase.class);
          private static final long serialVersionUID = -4115132557403913367L;
          private OutputCollector collector;
        
          @Override
          public void prepare(Map stormConf, TopologyContext context,
                    OutputCollector collector) {
               this.collector = collector;             
          }

          @Override
          public void execute(Tuple input) {
               String line = input.getString(0).trim();
               LOG.info("REALTIME: " + line);
               collector.ack(input);
          }

          @Override
          public void declareOutputFields(OutputFieldsDeclarer declarer) {
             
          }

     }

     public static void main(String[] args) throws AlreadyAliveException, InvalidTopologyException, InterruptedException {

          // Configure Kafka
          String zks = "h1:2181,h2:2181,h3:2181";
          String topic = "my-replicated-topic5";
          String zkRoot = "/storm"; // default zookeeper root configuration for storm
          String id = "word";
          BrokerHosts brokerHosts = new ZkHosts(zks);
          SpoutConfig spoutConf = new SpoutConfig(brokerHosts, topic, zkRoot, id);
          spoutConf.scheme = new SchemeAsMultiScheme(new StringScheme());
          spoutConf.forceFromStart = false;
          spoutConf.zkServers = Arrays.asList(new String[] {"h1", "h2", "h3"});
          spoutConf.zkPort = 2181;
        
          // Configure HDFS bolt
          RecordFormat format = new DelimitedRecordFormat()
                  .withFieldDelimiter("\t"); // use "\t" instead of "," for field delimiter
          SyncPolicy syncPolicy = new CountSyncPolicy(1000); // sync the filesystem after every 1k tuples
          FileRotationPolicy rotationPolicy = new TimedRotationPolicy(1.0f, TimeUnit.MINUTES); // rotate files
          FileNameFormat fileNameFormat = new DefaultFileNameFormat()
                  .withPath("/storm/").withPrefix("app_").withExtension(".log"); // set file name format
          HdfsBolt hdfsBolt = new HdfsBolt()
                  .withFsUrl("hdfs://h1:8020")
                  .withFileNameFormat(fileNameFormat)
                  .withRecordFormat(format)
                  .withRotationPolicy(rotationPolicy)
                  .withSyncPolicy(syncPolicy);
        
          // configure & build topology
          TopologyBuilder builder = new TopologyBuilder();
          builder.setSpout("kafka-reader", new KafkaSpout(spoutConf), 5);
          builder.setBolt("to-upper", new KafkaWordToUpperCase(), 3).shuffleGrouping("kafka-reader");
          builder.setBolt("hdfs-bolt", hdfsBolt, 2).shuffleGrouping("to-upper");
          builder.setBolt("realtime", new RealtimeBolt(), 2).shuffleGrouping("to-upper");
        
          // submit topology
          Config conf = new Config();
          String name = DistributeWordTopology.class.getSimpleName();
          if (args != null && args.length > 0) {
               String nimbus = args[0];
               conf.put(Config.NIMBUS_HOST, nimbus);
               conf.setNumWorkers(3);
               StormSubmitter.submitTopologyWithProgressBar(name, conf, builder.createTopology());
          } else {
               conf.setMaxTaskParallelism(3);
               LocalCluster cluster = new LocalCluster();
               cluster.submitTopology(name, conf, builder.createTopology());
               Thread.sleep(60000);
               cluster.shutdown();
          }
     }

}

上面代码中，名称为to-upper的Bolt将接收到的字符串行转换成大写以后，会将处理过的数据向后面的hdfs-bolt、realtime这两个Bolt各发一份拷贝，然后由这两个Bolt分别根据实际需要（实时/离线）单独处理。
打包后，在Storm集群上部署并运行这个Topology：

bin/storm jar ~/storm-examples-0.0.1-SNAPSHOT.jar org.shirdrn.storm.examples.DistributeWordTopology h1

可以通过Storm UI查看Topology运行情况，可以查看HDFS上生成的数据。