kafka分区策略

2017/03/14 10:16
阅读数 344
public class kafkaProducer  extends Thread{

    private String topic;  

    public kafkaProducer(String topic){  
        super();  
        this.topic = topic;  
    }  


    @Override  
    public void run() {  
        Producer producer = createProducer();  
        int i=0;  
        while(true){
            i++;
            String  string = "hello"+i;
            producer.send(new KeyedMessage<Integer, String>(topic,string));  
            if(i==100){
                break;
            }
            try {  
                TimeUnit.SECONDS.sleep(1);  
            } catch (InterruptedException e) {  
                e.printStackTrace();  
            }  
        }  
    }  

    private Producer createProducer() {  
        Properties properties = new Properties();  
        properties.put("serializer.class", StringEncoder.class.getName());  
        properties.put("metadata.broker.list", "ip1:9092,ip2:9092,ip3:9092");
        return new Producer<Integer, String>(new ProducerConfig(properties));  
     }  


    public static void main(String[] args) {  
        new kafkaProducer("user11").start();  
    } 
} 

发现其只向topic:user11中的某一个partiton中写数据。后来查了一下相关的参数:partitioner.class

partitioner.class

# 分区的策略
# 默认为kafka.producer.DefaultPartitioner,取模
partitioner.class = kafka.producer.DefaultPartitioner

在上面的程序中,我在producer中没有定义分区策略,也就是说程序采用默认的kafka.producer.DefaultPartitioner,来看看源码中是怎么定义的:

class DefaultPartitioner(props: VerifiableProperties = null) extends Partitioner {
  private val random = new java.util.Random

  def partition(key: Any, numPartitions: Int): Int = {
    Utils.abs(key.hashCode) % numPartitions
  }
}

其核心思想就是对每个消息的key的hash值对partition数取模得到。再来看看我的程序中有这么一段:

producer.send(new KeyedMessage<Integer, String>(topic,string))

来看看keyMessage:

case class KeyedMessage[K, V](val topic: String, val key: K, val partKey: Any, val message: V) {
  if(topic == null)
    throw new IllegalArgumentException("Topic cannot be null.")

  def this(topic: String, message: V) = this(topic, null.asInstanceOf[K], null, message)

  def this(topic: String, key: K, message: V) = this(topic, key, key, message)

  def partitionKey = {
    if(partKey != null)
      partKey
    else if(hasKey)
      key
    else
      null  
  }

  def hasKey = key != null
}

由于上面生产者代码中没有传入key,所以程序调用:

def this(topic: String, message: V) = this(topic, null.asInstanceOf[K], null, message)

这里的key值可以为空,在这种情况下, kafka会将这个消息发送到哪个分区上呢?依据Kafka官方的文档, 默认的分区类会随机挑选一个分区。这里的随机是指在参数"topic.metadata.refresh.ms"刷新后随机选择一个, 这个时间段内总是使用唯一的分区。 默认情况下每十分钟才可能重新选择一个新的分区。

private def getPartition(topic: String, key: Any, topicPartitionList: Seq[PartitionAndLeader]): Int = {
    val numPartitions = topicPartitionList.size
    if(numPartitions <= 0)
      throw new UnknownTopicOrPartitionException("Topic " + topic + " doesn't exist")
    val partition =
      if(key == null) {
        // If the key is null, we don't really need a partitioner
        // So we look up in the send partition cache for the topic to decide the target partition
        val id = sendPartitionPerTopicCache.get(topic)
        id match {
          case Some(partitionId) =>
            // directly return the partitionId without checking availability of the leader,
            // since we want to postpone the failure until the send operation anyways
            partitionId
          case None => 
            val availablePartitions = topicPartitionList.filter(_.leaderBrokerIdOpt.isDefined)
            if (availablePartitions.isEmpty)
              throw new LeaderNotAvailableException("No leader for any partition in topic " + topic)
            val index = Utils.abs(Random.nextInt) % availablePartitions.size
            val partitionId = availablePartitions(index).partitionId
            sendPartitionPerTopicCache.put(topic, partitionId)
            partitionId
        }
      } else
        partitioner.partition(key, numPartitions)
    if(partition < 0 || partition >= numPartitions)
      throw new UnknownTopicOrPartitionException("Invalid partition id: " + partition + " for topic " + topic +
        "; Valid values are in the inclusive range of [0, " + (numPartitions-1) + "]")
    trace("Assigning message of topic %s and key %s to a selected partition %d".format(topic, if (key == null) "[none]" else key.toString, partition))
    partition
  }

如果key为null, 它会从sendPartitionPerTopicCache查选缓存的分区, 如果没有,随机选择一个分区,否则就用缓存的分区。

LinkedIn工程师Guozhang Wang在邮件列表中解释了这一问题,
最初kafka是按照大部分用户理解的那样每次都随机选择一个分区, 后来改成了定期选择一个分区, 这是为了减少服务器段socket的数量。不过这的确很误导用户,据称0.8.2版本后又改回了每次随机选取。但是我查看0.8.2的代码还没看到改动。

所以,如果有可能,还是为KeyedMessage设置一个key值吧。比如:

producer.send(new KeyedMessage<String, String>(topic,String.valueOf(i),string)); 

自定义partitioner.class

如下所示为自定义的分区函数,分区函数实现了Partitioner接口

public class PersonalPartition implements Partitioner{

    public PersonalPartition(VerifiableProperties properties){

    }

    public int partition(Object arg0, int arg1) {
        if(arg0==null){
            return 0;
        }
        else{
            return 1;
        }   
    }

}

然后修改配置即可:

properties.put("partitioner.class", "com.xx.kafka.PersonalPartition"); 

向指定的partition写入数据

当然,也可以向topic中指定的partition中写数据,如下代码为向”user11”中partition 1中写入数据:

public class kafkaProducer  extends Thread{

    private String topic;  

    public kafkaProducer(String topic){  
        super();  
        this.topic = topic;  
    }  

    @Override  
    public void run() {  
        KafkaProducer producer = createProducer();  
        int i=0;  
        while(true){
            i++;
            String  string = "hello"+i;
            producer.send(new ProducerRecord(topic,1,null,string.getBytes()));
            if(i==10000){
                break;
            }
            try {  
                TimeUnit.SECONDS.sleep(1);  
            } catch (InterruptedException e) {  
                e.printStackTrace();  
            }  
        }  
    }  

    private KafkaProducer createProducer() {  
        Properties properties = new Properties();  
        properties.put("serializer.class", StringEncoder.class.getName());  
        properties.put("metadata.broker.list", "ip1:9092,ip2:9092,ip3:9092");
        return new KafkaProducer(properties);  
     } 
 } 

展开阅读全文
打赏
0
0 收藏
分享
加载中
更多评论
打赏
0 评论
0 收藏
0
分享
返回顶部
顶部