Skip to content

Latest commit

 

History

History
655 lines (493 loc) · 20.9 KB

Storm集成Redis详解.md

File metadata and controls

655 lines (493 loc) · 20.9 KB

Storm 集成 Redis 详解

一、简介
二、集成案例
三、storm-redis 实现原理
四、自定义RedisBolt实现词频统计

一、简介

Storm-Redis 提供了 Storm 与 Redis 的集成支持,你只需要引入对应的依赖即可使用:

<dependency>
    <groupId>org.apache.storm</groupId>
    <artifactId>storm-redis</artifactId>
    <version>${storm.version}</version>
    <type>jar</type>
</dependency> 

Storm-Redis 使用 Jedis 为 Redis 客户端,并提供了如下三个基本的 Bolt 实现:

  • RedisLookupBolt:从 Redis 中查询数据;
  • RedisStoreBolt:存储数据到 Redis;
  • RedisFilterBolt : 查询符合条件的数据;

RedisLookupBoltRedisStoreBoltRedisFilterBolt 均继承自 AbstractRedisBolt 抽象类。我们可以通过继承该抽象类,实现自定义 RedisBolt,进行功能的拓展。

二、集成案例

2.1 项目结构

这里首先给出一个集成案例:进行词频统计并将最后的结果存储到 Redis。项目结构如下:

用例源码下载地址:storm-redis-integration

2.2 项目依赖

项目主要依赖如下:

<properties>
    <storm.version>1.2.2</storm.version>
</properties>

<dependencies>
    <dependency>
        <groupId>org.apache.storm</groupId>
        <artifactId>storm-core</artifactId>
        <version>${storm.version}</version>
    </dependency>
    <dependency>
        <groupId>org.apache.storm</groupId>
        <artifactId>storm-redis</artifactId>
        <version>${storm.version}</version>
    </dependency>
</dependencies>

2.3 DataSourceSpout

/**
 * 产生词频样本的数据源
 */
public class DataSourceSpout extends BaseRichSpout {

    private List<String> list = Arrays.asList("Spark", "Hadoop", "HBase", "Storm", "Flink", "Hive");

    private SpoutOutputCollector spoutOutputCollector;

    @Override
    public void open(Map map, TopologyContext topologyContext, SpoutOutputCollector spoutOutputCollector) {
        this.spoutOutputCollector = spoutOutputCollector;
    }

    @Override
    public void nextTuple() {
        // 模拟产生数据
        String lineData = productData();
        spoutOutputCollector.emit(new Values(lineData));
        Utils.sleep(1000);
    }

    @Override
    public void declareOutputFields(OutputFieldsDeclarer outputFieldsDeclarer) {
        outputFieldsDeclarer.declare(new Fields("line"));
    }


    /**
     * 模拟数据
     */
    private String productData() {
        Collections.shuffle(list);
        Random random = new Random();
        int endIndex = random.nextInt(list.size()) % (list.size()) + 1;
        return StringUtils.join(list.toArray(), "\t", 0, endIndex);
    }

}

产生的模拟数据格式如下:

Spark	HBase
Hive	Flink	Storm	Hadoop	HBase	Spark
Flink
HBase	Storm
HBase	Hadoop	Hive	Flink
HBase	Flink	Hive	Storm
Hive	Flink	Hadoop
HBase	Hive
Hadoop	Spark	HBase	Storm

2.4 SplitBolt

/**
 * 将每行数据按照指定分隔符进行拆分
 */
public class SplitBolt extends BaseRichBolt {

    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.getStringByField("line");
        String[] words = line.split("\t");
        for (String word : words) {
            collector.emit(new Values(word, String.valueOf(1)));
        }
    }

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

2.5 CountBolt

/**
 * 进行词频统计
 */
public class CountBolt extends BaseRichBolt {

    private Map<String, Integer> counts = new HashMap<>();

    private OutputCollector collector;


    @Override
    public void prepare(Map stormConf, TopologyContext context, OutputCollector collector) {
            this.collector=collector;
    }

    @Override
    public void execute(Tuple input) {
        String word = input.getStringByField("word");
        Integer count = counts.get(word);
        if (count == null) {
            count = 0;
        }
        count++;
        counts.put(word, count);
        // 输出
        collector.emit(new Values(word, String.valueOf(count)));

    }

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

2.6 WordCountStoreMapper

实现 RedisStoreMapper 接口,定义 tuple 与 Redis 中数据的映射关系:即需要指定 tuple 中的哪个字段为 key,哪个字段为 value,并且存储到 Redis 的何种数据结构中。

/**
 * 定义 tuple 与 Redis 中数据的映射关系
 */
public class  WordCountStoreMapper implements RedisStoreMapper {
    private RedisDataTypeDescription description;
    private final String hashKey = "wordCount";

    public WordCountStoreMapper() {
        description = new RedisDataTypeDescription(
                RedisDataTypeDescription.RedisDataType.HASH, hashKey);
    }

    @Override
    public RedisDataTypeDescription getDataTypeDescription() {
        return description;
    }

    @Override
    public String getKeyFromTuple(ITuple tuple) {
        return tuple.getStringByField("word");
    }

    @Override
    public String getValueFromTuple(ITuple tuple) {
        return tuple.getStringByField("count");
    }
}

2.7 WordCountToRedisApp

/**
 * 进行词频统计 并将统计结果存储到 Redis 中
 */
public class WordCountToRedisApp {

    private static final String DATA_SOURCE_SPOUT = "dataSourceSpout";
    private static final String SPLIT_BOLT = "splitBolt";
    private static final String COUNT_BOLT = "countBolt";
    private static final String STORE_BOLT = "storeBolt";

    //在实际开发中这些参数可以将通过外部传入 使得程序更加灵活
    private static final String REDIS_HOST = "192.168.200.226";
    private static final int REDIS_PORT = 6379;

    public static void main(String[] args) {
        TopologyBuilder builder = new TopologyBuilder();
        builder.setSpout(DATA_SOURCE_SPOUT, new DataSourceSpout());
        // split
        builder.setBolt(SPLIT_BOLT, new SplitBolt()).shuffleGrouping(DATA_SOURCE_SPOUT);
        // count
        builder.setBolt(COUNT_BOLT, new CountBolt()).shuffleGrouping(SPLIT_BOLT);
        // save to redis
        JedisPoolConfig poolConfig = new JedisPoolConfig.Builder()
                .setHost(REDIS_HOST).setPort(REDIS_PORT).build();
        RedisStoreMapper storeMapper = new WordCountStoreMapper();
        RedisStoreBolt storeBolt = new RedisStoreBolt(poolConfig, storeMapper);
        builder.setBolt(STORE_BOLT, storeBolt).shuffleGrouping(COUNT_BOLT);

        // 如果外部传参 cluster 则代表线上环境启动否则代表本地启动
        if (args.length > 0 && args[0].equals("cluster")) {
            try {
                StormSubmitter.submitTopology("ClusterWordCountToRedisApp", new Config(), builder.createTopology());
            } catch (AlreadyAliveException | InvalidTopologyException | AuthorizationException e) {
                e.printStackTrace();
            }
        } else {
            LocalCluster cluster = new LocalCluster();
            cluster.submitTopology("LocalWordCountToRedisApp",
                    new Config(), builder.createTopology());
        }
    }
}

2.8 启动测试

可以用直接使用本地模式运行,也可以打包后提交到服务器集群运行。本仓库提供的源码默认采用 maven-shade-plugin 进行打包,打包命令如下:

# mvn clean package -D maven.test.skip=true

启动后,查看 Redis 中的数据:

三、storm-redis 实现原理

3.1 AbstractRedisBolt

RedisLookupBoltRedisStoreBoltRedisFilterBolt 均继承自 AbstractRedisBolt 抽象类,和我们自定义实现 Bolt 一样,AbstractRedisBolt 间接继承自 BaseRichBolt

AbstractRedisBolt 中比较重要的是 prepare 方法,在该方法中通过外部传入的 jedis 连接池配置 ( jedisPoolConfig/jedisClusterConfig) 创建用于管理 Jedis 实例的容器 JedisCommandsInstanceContainer

public abstract class AbstractRedisBolt extends BaseTickTupleAwareRichBolt {
    protected OutputCollector collector;

    private transient JedisCommandsInstanceContainer container;

    private JedisPoolConfig jedisPoolConfig;
    private JedisClusterConfig jedisClusterConfig;

   ......
   
    @Override
    public void prepare(Map map, TopologyContext topologyContext, OutputCollector collector) {
        // FIXME: stores map (stormConf), topologyContext and expose these to derived classes
        this.collector = collector;

        if (jedisPoolConfig != null) {
            this.container = JedisCommandsContainerBuilder.build(jedisPoolConfig);
        } else if (jedisClusterConfig != null) {
            this.container = JedisCommandsContainerBuilder.build(jedisClusterConfig);
        } else {
            throw new IllegalArgumentException("Jedis configuration not found");
        }
    }

  .......
}

JedisCommandsInstanceContainerbuild() 方法如下,实际上就是创建 JedisPool 或 JedisCluster 并传入容器中。

public static JedisCommandsInstanceContainer build(JedisPoolConfig config) {
        JedisPool jedisPool = new JedisPool(DEFAULT_POOL_CONFIG, config.getHost(), config.getPort(), config.getTimeout(), config.getPassword(), config.getDatabase());
        return new JedisContainer(jedisPool);
    }

 public static JedisCommandsInstanceContainer build(JedisClusterConfig config) {
        JedisCluster jedisCluster = new JedisCluster(config.getNodes(), config.getTimeout(), config.getTimeout(), config.getMaxRedirections(), config.getPassword(), DEFAULT_POOL_CONFIG);
        return new JedisClusterContainer(jedisCluster);
    }

3.2 RedisStoreBolt和RedisLookupBolt

RedisStoreBolt 中比较重要的是 process 方法,该方法主要从 storeMapper 中获取传入 key/value 的值,并按照其存储类型 dataType 调用 jedisCommand 的对应方法进行存储。

RedisLookupBolt 的实现基本类似,从 lookupMapper 中获取传入的 key 值,并进行查询操作。

public class RedisStoreBolt extends AbstractRedisBolt {
    private final RedisStoreMapper storeMapper;
    private final RedisDataTypeDescription.RedisDataType dataType;
    private final String additionalKey;

   public RedisStoreBolt(JedisPoolConfig config, RedisStoreMapper storeMapper) {
        super(config);
        this.storeMapper = storeMapper;

        RedisDataTypeDescription dataTypeDescription = storeMapper.getDataTypeDescription();
        this.dataType = dataTypeDescription.getDataType();
        this.additionalKey = dataTypeDescription.getAdditionalKey();
    }

    public RedisStoreBolt(JedisClusterConfig config, RedisStoreMapper storeMapper) {
        super(config);
        this.storeMapper = storeMapper;

        RedisDataTypeDescription dataTypeDescription = storeMapper.getDataTypeDescription();
        this.dataType = dataTypeDescription.getDataType();
        this.additionalKey = dataTypeDescription.getAdditionalKey();
    }
       
  
    @Override
    public void process(Tuple input) {
        String key = storeMapper.getKeyFromTuple(input);
        String value = storeMapper.getValueFromTuple(input);

        JedisCommands jedisCommand = null;
        try {
            jedisCommand = getInstance();

            switch (dataType) {
                case STRING:
                    jedisCommand.set(key, value);
                    break;

                case LIST:
                    jedisCommand.rpush(key, value);
                    break;

                case HASH:
                    jedisCommand.hset(additionalKey, key, value);
                    break;

                case SET:
                    jedisCommand.sadd(key, value);
                    break;

                case SORTED_SET:
                    jedisCommand.zadd(additionalKey, Double.valueOf(value), key);
                    break;

                case HYPER_LOG_LOG:
                    jedisCommand.pfadd(key, value);
                    break;

                case GEO:
                    String[] array = value.split(":");
                    if (array.length != 2) {
                        throw new IllegalArgumentException("value structure should be longitude:latitude");
                    }

                    double longitude = Double.valueOf(array[0]);
                    double latitude = Double.valueOf(array[1]);
                    jedisCommand.geoadd(additionalKey, longitude, latitude, key);
                    break;

                default:
                    throw new IllegalArgumentException("Cannot process such data type: " + dataType);
            }

            collector.ack(input);
        } catch (Exception e) {
            this.collector.reportError(e);
            this.collector.fail(input);
        } finally {
            returnInstance(jedisCommand);
        }
    }

     .........
}

3.3 JedisCommands

JedisCommands 接口中定义了所有的 Redis 客户端命令,它有以下三个实现类,分别是 Jedis、JedisCluster、ShardedJedis。Strom 中主要使用前两种实现类,具体调用哪一个实现类来执行命令,由传入的是 jedisPoolConfig 还是 jedisClusterConfig 来决定。

3.4 RedisMapper 和 TupleMapper

RedisMapper 和 TupleMapper 定义了 tuple 和 Redis 中的数据如何进行映射转换。

1. TupleMapper

TupleMapper 主要定义了两个方法:

  • getKeyFromTuple(ITuple tuple): 从 tuple 中获取那个字段作为 Key;

  • getValueFromTuple(ITuple tuple):从 tuple 中获取那个字段作为 Value;

2. RedisMapper

定义了获取数据类型的方法 getDataTypeDescription(),RedisDataTypeDescription 中 RedisDataType 枚举类定义了所有可用的 Redis 数据类型:

public class RedisDataTypeDescription implements Serializable { 

    public enum RedisDataType { STRING, HASH, LIST, SET, SORTED_SET, HYPER_LOG_LOG, GEO }
     ......
    }

3. RedisStoreMapper

RedisStoreMapper 继承 TupleMapper 和 RedisMapper 接口,用于数据存储时,没有定义额外方法。

4. RedisLookupMapper

RedisLookupMapper 继承 TupleMapper 和 RedisMapper 接口:

  • 定义了 declareOutputFields 方法,声明输出的字段。
  • 定义了 toTuple 方法,将查询结果组装为 Storm 的 Values 的集合,并用于发送。

下面的例子表示从输入 Tuple 的获取 word 字段作为 key,使用 RedisLookupBolt 进行查询后,将 key 和查询结果 value 组装为 values 并发送到下一个处理单元。

class WordCountRedisLookupMapper implements RedisLookupMapper {
    private RedisDataTypeDescription description;
    private final String hashKey = "wordCount";

    public WordCountRedisLookupMapper() {
        description = new RedisDataTypeDescription(
                RedisDataTypeDescription.RedisDataType.HASH, hashKey);
    }

    @Override
    public List<Values> toTuple(ITuple input, Object value) {
        String member = getKeyFromTuple(input);
        List<Values> values = Lists.newArrayList();
        values.add(new Values(member, value));
        return values;
    }

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

    @Override
    public RedisDataTypeDescription getDataTypeDescription() {
        return description;
    }

    @Override
    public String getKeyFromTuple(ITuple tuple) {
        return tuple.getStringByField("word");
    }

    @Override
    public String getValueFromTuple(ITuple tuple) {
        return null;
    }
} 

5. RedisFilterMapper

RedisFilterMapper 继承 TupleMapper 和 RedisMapper 接口,用于查询数据时,定义了 declareOutputFields 方法,声明输出的字段。如下面的实现:

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

四、自定义RedisBolt实现词频统计

4.1 实现原理

自定义 RedisBolt:主要利用 Redis 中哈希结构的 hincrby key field 命令进行词频统计。在 Redis 中 hincrby 的执行效果如下。hincrby 可以将字段按照指定的值进行递增,如果该字段不存在的话,还会新建该字段,并赋值为 0。通过这个命令可以非常轻松的实现词频统计功能。

redis>  HSET myhash field 5
(integer) 1
redis>  HINCRBY myhash field 1
(integer) 6
redis>  HINCRBY myhash field -1
(integer) 5
redis>  HINCRBY myhash field -10
(integer) -5
redis> 

4.2 项目结构

4.3 自定义RedisBolt的代码实现

/**
 * 自定义 RedisBolt 利用 Redis 的哈希数据结构的 hincrby key field 命令进行词频统计
 */
public class RedisCountStoreBolt extends AbstractRedisBolt {

    private final RedisStoreMapper storeMapper;
    private final RedisDataTypeDescription.RedisDataType dataType;
    private final String additionalKey;

    public RedisCountStoreBolt(JedisPoolConfig config, RedisStoreMapper storeMapper) {
        super(config);
        this.storeMapper = storeMapper;
        RedisDataTypeDescription dataTypeDescription = storeMapper.getDataTypeDescription();
        this.dataType = dataTypeDescription.getDataType();
        this.additionalKey = dataTypeDescription.getAdditionalKey();
    }

    @Override
    protected void process(Tuple tuple) {
        String key = storeMapper.getKeyFromTuple(tuple);
        String value = storeMapper.getValueFromTuple(tuple);

        JedisCommands jedisCommand = null;
        try {
            jedisCommand = getInstance();
            if (dataType == RedisDataTypeDescription.RedisDataType.HASH) {
                jedisCommand.hincrBy(additionalKey, key, Long.valueOf(value));
            } else {
                throw new IllegalArgumentException("Cannot process such data type for Count: " + dataType);
            }

            collector.ack(tuple);
        } catch (Exception e) {
            this.collector.reportError(e);
            this.collector.fail(tuple);
        } finally {
            returnInstance(jedisCommand);
        }
    }

    @Override
    public void declareOutputFields(OutputFieldsDeclarer declarer) {

    }
}

4.4 CustomRedisCountApp

/**
 * 利用自定义的 RedisBolt 实现词频统计
 */
public class CustomRedisCountApp {

    private static final String DATA_SOURCE_SPOUT = "dataSourceSpout";
    private static final String SPLIT_BOLT = "splitBolt";
    private static final String STORE_BOLT = "storeBolt";

    private static final String REDIS_HOST = "192.168.200.226";
    private static final int REDIS_PORT = 6379;

    public static void main(String[] args) {
        TopologyBuilder builder = new TopologyBuilder();
        builder.setSpout(DATA_SOURCE_SPOUT, new DataSourceSpout());
        // split
        builder.setBolt(SPLIT_BOLT, new SplitBolt()).shuffleGrouping(DATA_SOURCE_SPOUT);
        // save to redis and count
        JedisPoolConfig poolConfig = new JedisPoolConfig.Builder()
                .setHost(REDIS_HOST).setPort(REDIS_PORT).build();
        RedisStoreMapper storeMapper = new WordCountStoreMapper();
        RedisCountStoreBolt countStoreBolt = new RedisCountStoreBolt(poolConfig, storeMapper);
        builder.setBolt(STORE_BOLT, countStoreBolt).shuffleGrouping(SPLIT_BOLT);

        // 如果外部传参 cluster 则代表线上环境启动,否则代表本地启动
        if (args.length > 0 && args[0].equals("cluster")) {
            try {
                StormSubmitter.submitTopology("ClusterCustomRedisCountApp", new Config(), builder.createTopology());
            } catch (AlreadyAliveException | InvalidTopologyException | AuthorizationException e) {
                e.printStackTrace();
            }
        } else {
            LocalCluster cluster = new LocalCluster();
            cluster.submitTopology("LocalCustomRedisCountApp",
                    new Config(), builder.createTopology());
        }
    }
}

参考资料

  1. Storm Redis Integration