Kafka+Storm+HDFS整合实践

在基于Hadoop平台的很多应用场景中,我们需要对数据进行离线和实时分析,离线分析可以很容易地借助于Hive来实现统计分析,但是对于实时的需求Hive就不合适了。实时应用场景可以使用Storm,它是一个实时处理系统,它为实时处理类应用提供了一个计算模型,可以很容易地进行编程处理。为了统一离线和实时计算,一般情况下,我们都希望将离线和实时计算的数据源的集合统一起来作为输入,然后将数据的流向分别经由实时系统和离线分析系统,分别进行分析处理,这时我们可以考虑将数据源(如使用Flume收集日志)直接连接一个消息中间件,如Kafka,可以整合Flume+Kafka,Flume作为消息的Producer,生产的消息数据(日志数据、业务请求数据等等)发布到Kafka中,然后通过订阅的方式,使用Storm的Topology作为消息的Consumer,在Storm集群中分别进行如下两个需求场景的处理:

  • 直接使用Storm的Topology对数据进行实时分析处理
  • 整合Storm+HDFS,将消息处理后写入HDFS进行离线分析处理

实时处理,只要开发满足业务需要的Topology即可,不做过多说明。这里,我们主要从安装配置Kafka、Storm,以及整合Kafka+Storm、整合Storm+HDFS、整合Kafka+Storm+HDFS这几点来配置实践,满足上面提出的一些需求。配置实践使用的软件包如下所示:

  • zookeeper-3.4.5.tar.gz
  • kafka_2.9.2-0.8.1.1.tgz
  • apache-storm-0.9.2-incubating.tar.gz
  • hadoop-2.2.0.tar.gz

程序配置运行所基于的操作系统为CentOS 5.11。

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/
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/
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/
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:

01 <dependency>
02      <groupId>org.apache.storm</groupId>
03      <artifactId>storm-core</artifactId>
04      <version>0.9.2-incubating</version>
05      <scope>provided</scope>
06 </dependency>
07 <dependency>
08      <groupId>org.apache.storm</groupId>
09      <artifactId>storm-kafka</artifactId>
10      <version>0.9.2-incubating</version>
11 </dependency>
12 <dependency>
13      <groupId>org.apache.kafka</groupId>
14      <artifactId>kafka_2.9.2</artifactId>
15      <version>0.8.1.1</version>
16      <exclusions>
17           <exclusion>
18                <groupId>org.apache.zookeeper</groupId>
19                <artifactId>zookeeper</artifactId>
20           </exclusion>
21           <exclusion>
22                <groupId>log4j</groupId>
23                <artifactId>log4j</artifactId>
24           </exclusion>
25      </exclusions>
26 </dependency>

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:

001 package org.shirdrn.storm.examples;
002  
003 import java.util.Arrays;
004 import java.util.HashMap;
005 import java.util.Iterator;
006 import java.util.Map;
007 import java.util.Map.Entry;
008 import java.util.concurrent.atomic.AtomicInteger;
009  
010 import org.apache.commons.logging.Log;
011 import org.apache.commons.logging.LogFactory;
012  
013 import storm.kafka.BrokerHosts;
014 import storm.kafka.KafkaSpout;
015 import storm.kafka.SpoutConfig;
016 import storm.kafka.StringScheme;
017 import storm.kafka.ZkHosts;
018 import backtype.storm.Config;
019 import backtype.storm.LocalCluster;
020 import backtype.storm.StormSubmitter;
021 import backtype.storm.generated.AlreadyAliveException;
022 import backtype.storm.generated.InvalidTopologyException;
023 import backtype.storm.spout.SchemeAsMultiScheme;
024 import backtype.storm.task.OutputCollector;
025 import backtype.storm.task.TopologyContext;
026 import backtype.storm.topology.OutputFieldsDeclarer;
027 import backtype.storm.topology.TopologyBuilder;
028 import backtype.storm.topology.base.BaseRichBolt;
029 import backtype.storm.tuple.Fields;
030 import backtype.storm.tuple.Tuple;
031 import backtype.storm.tuple.Values;
032  
033 public class MyKafkaTopology {
034  
035      public static class KafkaWordSplitter extends BaseRichBolt {
036  
037           private static final Log LOG = LogFactory.getLog(KafkaWordSplitter.class);
038           private static final long serialVersionUID = 886149197481637894L;
039           private OutputCollector collector;
040           
041           @Override
042           public void prepare(Map stormConf, TopologyContext context,
043                     OutputCollector collector) {
044                this.collector = collector;             
045           }
046  
047           @Override
048           public void execute(Tuple input) {
049                String line = input.getString(0);
050                LOG.info("RECV[kafka -> splitter] " + line);
051                String[] words = line.split("\\s+");
052                for(String word : words) {
053                     LOG.info("EMIT[splitter -> counter] " + word);
054                     collector.emit(input, new Values(word, 1));
055                }
056                collector.ack(input);
057           }
058  
059           @Override
060           public void declareOutputFields(OutputFieldsDeclarer declarer) {
061                declarer.declare(new Fields("word""count"));        
062           }
063           
064      }
065      
066      public static class WordCounter extends BaseRichBolt {
067  
068           private static final Log LOG = LogFactory.getLog(WordCounter.class);
069           private static final long serialVersionUID = 886149197481637894L;
070           private OutputCollector collector;
071           private Map<String, AtomicInteger> counterMap;
072           
073           @Override
074           public void prepare(Map stormConf, TopologyContext context,
075                     OutputCollector collector) {
076                this.collector = collector;   
077                this.counterMap = new HashMap<String, AtomicInteger>();
078           }
079  
080           @Override
081           public void execute(Tuple input) {
082                String word = input.getString(0);
083                int count = input.getInteger(1);
084                LOG.info("RECV[splitter -> counter] " + word + " : " + count);
085                AtomicInteger ai = this.counterMap.get(word);
086                if(ai == null) {
087                     ai = new AtomicInteger();
088                     this.counterMap.put(word, ai);
089                }
090                ai.addAndGet(count);
091                collector.ack(input);
092                LOG.info("CHECK statistics map: " this.counterMap);
093           }
094  
095           @Override
096           public void cleanup() {
097                LOG.info("The final result:");
098                Iterator<Entry<String, AtomicInteger>> iter =this.counterMap.entrySet().iterator();
099                while(iter.hasNext()) {
100                     Entry<String, AtomicInteger> entry = iter.next();
101                     LOG.info(entry.getKey() + "\t:\t" + entry.getValue().get());
102                }
103                
104           }
105  
106           @Override
107           public void declareOutputFields(OutputFieldsDeclarer declarer) {
108                declarer.declare(new Fields("word""count"));        
109           }
110      }
111      
112      public static void main(String[] args) throws AlreadyAliveException, InvalidTopologyException, InterruptedException {
113           String zks = "h1:2181,h2:2181,h3:2181";
114           String topic = "my-replicated-topic5";
115           String zkRoot = "/storm"// default zookeeper root configuration for storm
116           String id = "word";
117           
118           BrokerHosts brokerHosts = new ZkHosts(zks);
119           SpoutConfig spoutConf = new SpoutConfig(brokerHosts, topic, zkRoot, id);
120           spoutConf.scheme = new SchemeAsMultiScheme(new StringScheme());
121           spoutConf.forceFromStart = false;
122           spoutConf.zkServers = Arrays.asList(new String[] {"h1""h2""h3"});
123           spoutConf.zkPort = 2181;
124           
125           TopologyBuilder builder = new TopologyBuilder();
126           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
127           builder.setBolt("word-splitter"new KafkaWordSplitter(),2).shuffleGrouping("kafka-reader");
128           builder.setBolt("word-counter"new WordCounter()).fieldsGrouping("word-splitter"new Fields("word"));
129           
130           Config conf = new Config();
131           
132           String name = MyKafkaTopology.class.getSimpleName();
133           if (args != null && args.length > 0) {
134                // Nimbus host name passed from command line
135                conf.put(Config.NIMBUS_HOST, args[0]);
136                conf.setNumWorkers(3);
137                StormSubmitter.submitTopologyWithProgressBar(name, conf, builder.createTopology());
138           else {
139                conf.setMaxTaskParallelism(3);
140                LocalCluster cluster = new LocalCluster();
141                cluster.submitTopology(name, conf, builder.createTopology());
142                Thread.sleep(60000);
143                cluster.shutdown();
144           }
145      }
146 }

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:

1 cp /usr/local/kafka/libs/kafka_2.9.2-0.8.1.1.jar /usr/local/storm/lib/
2 cp /usr/local/kafka/libs/scala-library-2.9.2.jar /usr/local/storm/lib/
3 cp /usr/local/kafka/libs/metrics-core-2.2.0.jar /usr/local/storm/lib/
4 cp /usr/local/kafka/libs/snappy-java-1.0.5.jar /usr/local/storm/lib/
5 cp /usr/local/kafka/libs/zkclient-0.3.jar /usr/local/storm/lib/
6 cp /usr/local/kafka/libs/log4j-1.2.15.jar /usr/local/storm/lib/
7 cp /usr/local/kafka/libs/slf4j-api-1.7.2.jar /usr/local/storm/lib/
8 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:

1 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:

1 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:

001 package org.shirdrn.storm.examples;
002  
003 import java.text.DateFormat;
004 import java.text.SimpleDateFormat;
005 import java.util.Date;
006 import java.util.Map;
007 import java.util.Random;
008  
009 import org.apache.commons.logging.Log;
010 import org.apache.commons.logging.LogFactory;
011 import org.apache.storm.hdfs.bolt.HdfsBolt;
012 import org.apache.storm.hdfs.bolt.format.DefaultFileNameFormat;
013 import org.apache.storm.hdfs.bolt.format.DelimitedRecordFormat;
014 import org.apache.storm.hdfs.bolt.format.FileNameFormat;
015 import org.apache.storm.hdfs.bolt.format.RecordFormat;
016 import org.apache.storm.hdfs.bolt.rotation.FileRotationPolicy;
017 import org.apache.storm.hdfs.bolt.rotation.TimedRotationPolicy;
018 import org.apache.storm.hdfs.bolt.rotation.TimedRotationPolicy.TimeUnit;
019 import org.apache.storm.hdfs.bolt.sync.CountSyncPolicy;
020 import org.apache.storm.hdfs.bolt.sync.SyncPolicy;
021  
022 import backtype.storm.Config;
023 import backtype.storm.LocalCluster;
024 import backtype.storm.StormSubmitter;
025 import backtype.storm.generated.AlreadyAliveException;
026 import backtype.storm.generated.InvalidTopologyException;
027 import backtype.storm.spout.SpoutOutputCollector;
028 import backtype.storm.task.TopologyContext;
029 import backtype.storm.topology.OutputFieldsDeclarer;
030 import backtype.storm.topology.TopologyBuilder;
031 import backtype.storm.topology.base.BaseRichSpout;
032 import backtype.storm.tuple.Fields;
033 import backtype.storm.tuple.Values;
034 import backtype.storm.utils.Utils;
035  
036 public class StormToHDFSTopology {
037  
038      public static class EventSpout extends BaseRichSpout {
039  
040           private static final Log LOG = LogFactory.getLog(EventSpout.class);
041           private static final long serialVersionUID = 886149197481637894L;
042           private SpoutOutputCollector collector;
043           private Random rand;
044           private String[] records;
045           
046           @Override
047           public void open(Map conf, TopologyContext context,
048                     SpoutOutputCollector collector) {
049                this.collector = collector;   
050                rand = new Random();
051                records = new String[] {
052                          "10001     ef2da82d4c8b49c44199655dc14f39f6     4.2.1     HUAWEI G610-U00     HUAWEI     2     70:72:3c:73:8b:22     2014-10-13 12:36:35",
053                          "10001     ffb52739a29348a67952e47c12da54ef     4.3     GT-I9300     samsung     2     50:CC:F8:E4:22:E2     2014-10-13 12:36:02",
054                          "10001     ef2da82d4c8b49c44199655dc14f39f6     4.2.1     HUAWEI G610-U00     HUAWEI     2     70:72:3c:73:8b:22     2014-10-13 12:36:35"
055                };
056           }
057  
058  
059           @Override
060           public void nextTuple() {
061                Utils.sleep(1000);
062                DateFormat df = new SimpleDateFormat("yyyy-MM-dd_HH-mm-ss");
063                Date d = new Date(System.currentTimeMillis());
064                String minute = df.format(d);
065                String record = records[rand.nextInt(records.length)];
066                LOG.info("EMIT[spout -> hdfs] " + minute + " : " + record);
067                collector.emit(new Values(minute, record));
068           }
069  
070           @Override
071           public void declareOutputFields(OutputFieldsDeclarer declarer) {
072                declarer.declare(new Fields("minute""record"));        
073           }
074  
075  
076      }
077      
078      public static void main(String[] args) throws AlreadyAliveException, InvalidTopologyException, InterruptedException {
079           // use "|" instead of "," for field delimiter
080           RecordFormat format = new DelimitedRecordFormat()
081                   .withFieldDelimiter(" : ");
082  
083           // sync the filesystem after every 1k tuples
084           SyncPolicy syncPolicy = new CountSyncPolicy(1000);
085  
086           // rotate files
087           FileRotationPolicy rotationPolicy = new TimedRotationPolicy(1.0f, TimeUnit.MINUTES);
088  
089           FileNameFormat fileNameFormat = new DefaultFileNameFormat()
090                   .withPath("/storm/").withPrefix("app_").withExtension(".log");
091  
092           HdfsBolt hdfsBolt = new HdfsBolt()
093                   .withFsUrl("hdfs://h1:8020")
094                   .withFileNameFormat(fileNameFormat)
095                   .withRecordFormat(format)
096                   .withRotationPolicy(rotationPolicy)
097                   .withSyncPolicy(syncPolicy);
098           
099           TopologyBuilder builder = new TopologyBuilder();
100           builder.setSpout("event-spout"new EventSpout(), 3);
101           builder.setBolt("hdfs-bolt", hdfsBolt, 2).fieldsGrouping("event-spout"newFields("minute"));
102           
103           Config conf = new Config();
104           
105           String name = StormToHDFSTopology.class.getSimpleName();
106           if (args != null && args.length > 0) {
107                conf.put(Config.NIMBUS_HOST, args[0]);
108                conf.setNumWorkers(3);
109                StormSubmitter.submitTopologyWithProgressBar(name, conf, builder.createTopology());
110           else {
111                conf.setMaxTaskParallelism(3);
112                LocalCluster cluster = new LocalCluster();
113                cluster.submitTopology(name, conf, builder.createTopology());
114                Thread.sleep(60000);
115                cluster.shutdown();
116           }
117      }
118  
119 }

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:

01 <plugin>
02     <groupId>org.apache.maven.plugins</groupId>
03     <artifactId>maven-shade-plugin</artifactId>
04     <version>1.4</version>
05     <configuration>
06         <createDependencyReducedPom>true</createDependencyReducedPom>
07     </configuration>
08     <executions>
09         <execution>
10             <phase>package</phase>
11             <goals>
12                 <goal>shade</goal>
13             </goals>
14             <configuration>
15                 <transformers>
16                     <transformer
17                             implementation="org.apache.maven.plugins.shade.resource.ServicesResourceTransformer"/>
18                     <transformer
19                             implementation="org.apache.maven.plugins.shade.resource.ManifestResourceTransformer">
20                         <mainClass></mainClass>
21                     </transformer>
22                 </transformers>
23             </configuration>
24         </execution>
25     </executions>
26 </plugin>

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:

001 package org.shirdrn.storm.examples;
002  
003 import java.util.Arrays;
004 import java.util.Map;
005  
006 import org.apache.commons.logging.Log;
007 import org.apache.commons.logging.LogFactory;
008 import org.apache.storm.hdfs.bolt.HdfsBolt;
009 import org.apache.storm.hdfs.bolt.format.DefaultFileNameFormat;
010 import org.apache.storm.hdfs.bolt.format.DelimitedRecordFormat;
011 import org.apache.storm.hdfs.bolt.format.FileNameFormat;
012 import org.apache.storm.hdfs.bolt.format.RecordFormat;
013 import org.apache.storm.hdfs.bolt.rotation.FileRotationPolicy;
014 import org.apache.storm.hdfs.bolt.rotation.TimedRotationPolicy;
015 import org.apache.storm.hdfs.bolt.rotation.TimedRotationPolicy.TimeUnit;
016 import org.apache.storm.hdfs.bolt.sync.CountSyncPolicy;
017 import org.apache.storm.hdfs.bolt.sync.SyncPolicy;
018  
019 import storm.kafka.BrokerHosts;
020 import storm.kafka.KafkaSpout;
021 import storm.kafka.SpoutConfig;
022 import storm.kafka.StringScheme;
023 import storm.kafka.ZkHosts;
024 import backtype.storm.Config;
025 import backtype.storm.LocalCluster;
026 import backtype.storm.StormSubmitter;
027 import backtype.storm.generated.AlreadyAliveException;
028 import backtype.storm.generated.InvalidTopologyException;
029 import backtype.storm.spout.SchemeAsMultiScheme;
030 import backtype.storm.task.OutputCollector;
031 import backtype.storm.task.TopologyContext;
032 import backtype.storm.topology.OutputFieldsDeclarer;
033 import backtype.storm.topology.TopologyBuilder;
034 import backtype.storm.topology.base.BaseRichBolt;
035 import backtype.storm.tuple.Fields;
036 import backtype.storm.tuple.Tuple;
037 import backtype.storm.tuple.Values;
038  
039 public class DistributeWordTopology {
040      
041      public static class KafkaWordToUpperCase extends BaseRichBolt {
042  
043           private static final Log LOG = LogFactory.getLog(KafkaWordToUpperCase.class);
044           private static final long serialVersionUID = -5207232012035109026L;
045           private OutputCollector collector;
046           
047           @Override
048           public void prepare(Map stormConf, TopologyContext context,
049                     OutputCollector collector) {
050                this.collector = collector;             
051           }
052  
053           @Override
054           public void execute(Tuple input) {
055                String line = input.getString(0).trim();
056                LOG.info("RECV[kafka -> splitter] " + line);
057                if(!line.isEmpty()) {
058                     String upperLine = line.toUpperCase();
059                     LOG.info("EMIT[splitter -> counter] " + upperLine);
060                     collector.emit(input, new Values(upperLine, upperLine.length()));
061                }
062                collector.ack(input);
063           }
064  
065           @Override
066           public void declareOutputFields(OutputFieldsDeclarer declarer) {
067                declarer.declare(new Fields("line""len"));        
068           }
069           
070      }
071      
072      public static class RealtimeBolt extends BaseRichBolt {
073  
074           private static final Log LOG = LogFactory.getLog(KafkaWordToUpperCase.class);
075           private static final long serialVersionUID = -4115132557403913367L;
076           private OutputCollector collector;
077           
078           @Override
079           public void prepare(Map stormConf, TopologyContext context,
080                     OutputCollector collector) {
081                this.collector = collector;             
082           }
083  
084           @Override
085           public void execute(Tuple input) {
086                String line = input.getString(0).trim();
087                LOG.info("REALTIME: " + line);
088                collector.ack(input);
089           }
090  
091           @Override
092           public void declareOutputFields(OutputFieldsDeclarer declarer) {
093                
094           }
095  
096      }
097  
098      public static void main(String[] args) throws AlreadyAliveException, InvalidTopologyException, InterruptedException {
099  
100           // Configure Kafka
101           String zks = "h1:2181,h2:2181,h3:2181";
102           String topic = "my-replicated-topic5";
103           String zkRoot = "/storm"// default zookeeper root configuration for storm
104           String id = "word";
105           BrokerHosts brokerHosts = new ZkHosts(zks);
106           SpoutConfig spoutConf = new SpoutConfig(brokerHosts, topic, zkRoot, id);
107           spoutConf.scheme = new SchemeAsMultiScheme(new StringScheme());
108           spoutConf.forceFromStart = false;
109           spoutConf.zkServers = Arrays.asList(new String[] {"h1""h2""h3"});
110           spoutConf.zkPort = 2181;
111           
112           // Configure HDFS bolt
113           RecordFormat format = new DelimitedRecordFormat()
114                   .withFieldDelimiter("\t"); // use "\t" instead of "," for field delimiter
115           SyncPolicy syncPolicy = new CountSyncPolicy(1000); // sync the filesystem after every 1k tuples
116           FileRotationPolicy rotationPolicy = new TimedRotationPolicy(1.0f, TimeUnit.MINUTES); // rotate files
117           FileNameFormat fileNameFormat = new DefaultFileNameFormat()
118                   .withPath("/storm/").withPrefix("app_").withExtension(".log"); // set file name format
119           HdfsBolt hdfsBolt = new HdfsBolt()
120                   .withFsUrl("hdfs://h1:8020")
121                   .withFileNameFormat(fileNameFormat)
122                   .withRecordFormat(format)
123                   .withRotationPolicy(rotationPolicy)
124                   .withSyncPolicy(syncPolicy);
125           
126           // configure & build topology
127           TopologyBuilder builder = new TopologyBuilder();
128           builder.setSpout("kafka-reader"new KafkaSpout(spoutConf), 5);
129           builder.setBolt("to-upper"new KafkaWordToUpperCase(),3).shuffleGrouping("kafka-reader");
130           builder.setBolt("hdfs-bolt", hdfsBolt, 2).shuffleGrouping("to-upper");
131           builder.setBolt("realtime"new RealtimeBolt(), 2).shuffleGrouping("to-upper");
132           
133           // submit topology
134           Config conf = new Config();
135           String name = DistributeWordTopology.class.getSimpleName();
136           if (args != null && args.length > 0) {
137                String nimbus = args[0];
138                conf.put(Config.NIMBUS_HOST, nimbus);
139                conf.setNumWorkers(3);
140                StormSubmitter.submitTopologyWithProgressBar(name, conf, builder.createTopology());
141           else {
142                conf.setMaxTaskParallelism(3);
143                LocalCluster cluster = new LocalCluster();
144                cluster.submitTopology(name, conf, builder.createTopology());
145                Thread.sleep(60000);
146                cluster.shutdown();
147           }
148      }
149  
150 }

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:

1 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。

上一篇:彪悍开源的分析数据库-ClickHouse


下一篇:spark读取 kafka nginx网站日志消息 并写入HDFS中(转)