开篇导语
Flink将数据sink至Kafka的过程中,在初始化生产者对象FlinkKafkaProducer时通常会采用默认的分区器和序列化器,这样数据只会发送至指定Topic的某一个分区中。对于存在多分区的Topic我们一般要自定义分区器和序列化器,指定数据发送至不同分区的逻辑。
此篇博客所涉及的组件版本
Flink:1.10.0
Kafka:2.3.0
序列化器
在Kafka生产者将数据写入至Kafka集群中时,为了能够在网络中传输数据对象,需要先将数据进行序列化处理,对于初学者来说,在初始化生产者对象时,一般都会采用默认的序列化器。
默认的序列化器不会对数据进行任何操作,也不会生成key。如果我们需要指定数据的key或者在数据发送前进行一些定制化的操作,那么我们就需要自定义序列化器,并且在初始化生产者对象时指定我们自己的序列化器。
分区器
对于Kakfa中一个topic存在多个分区的情况下,我们怎么知道发送的数据会被分配到哪个分区呢,这时候就要通过分区器来进行区分。
在Kafka中,主要有以下四种数据分区策略
第一种分区策略:给定了分区号,直接将数据发送到指定的分区里面去
第二种分区策略:没有给定分区号,给定数据的key值,通过key取hashCode进行分区
第三种分区策略:既没有给定分区号,也没有给定key值,直接轮询进行分区
第四种分区策略:自定义分区
分区器就是以上分区策略的代码实现。
Flink中的Kafka序列化器
源码解读
在之前的Flink版中中,自定义Kafka序列化器都是实现KeyedSerializationSchema接口,看一下它的源码:
//表示当前接口已经不推荐使用 @Deprecated @PublicEvolving //当前接口实现时需要指定生产者所要传输的对象类型 public interface KeyedSerializationSchema<T> extends Serializable { //根据传入的对象生成key并序列化为字节数组 //如果没有生成key,这个方法可返回null。 byte[] serializeKey(T element); //根据传入的对象生成value并序列化为字节数组 byte[] serializeValue(T element); //根据传入的对象指定需要发送的Topic //此方法可以返回null,因为在初始化生产者对象的时候就已经指定了Topic。 String getTargetTopic(T element); }
以上接口存在三个方法,但是每个输入的参数都是一样的,代码复用性低。
于是现在的Flink版本一般推荐实现KafkaSerializationSchema接口来实现序列化器,看一下它的源码:
//当前接口实现时需要指定生产者所要传输的对象类型 @PublicEvolving public interface KafkaSerializationSchema<T> extends Serializable { /** * Serializes given element and returns it as a {@link ProducerRecord}. * * @param element element to be serialized * @param timestamp timestamp (can be null) * @return Kafka {@link ProducerRecord} */ ProducerRecord<byte[], byte[]> serialize(T element, @Nullable Long timestamp); }
可以看到当前接口只需要实现serialize方法,根据传入的对象构造并返回一个ProducerRecord对象。
我们来看一下ProducerRecord类的源码:
package org.apache.kafka.clients.producer; import java.util.Objects; import org.apache.kafka.common.header.Header; import org.apache.kafka.common.header.Headers; import org.apache.kafka.common.header.internals.RecordHeaders; public class ProducerRecord<K, V> { private final String topic; private final Integer partition; private final Headers headers; private final K key; private final V value; private final Long timestamp; public ProducerRecord(String topic, Integer partition, Long timestamp, K key, V value, Iterable<Header> headers) { if (topic == null) { throw new IllegalArgumentException("Topic cannot be null."); } else if (timestamp != null && timestamp < 0L) { throw new IllegalArgumentException(String.format("Invalid timestamp: %d. Timestamp should always be non-negative or null.", timestamp)); } else if (partition != null && partition < 0) { throw new IllegalArgumentException(String.format("Invalid partition: %d. Partition number should always be non-negative or null.", partition)); } else { this.topic = topic; this.partition = partition; this.key = key; this.value = value; this.timestamp = timestamp; this.headers = new RecordHeaders(headers); } } public ProducerRecord(String topic, Integer partition, Long timestamp, K key, V value) { this(topic, partition, timestamp, key, value, (Iterable)null); } public ProducerRecord(String topic, Integer partition, K key, V value, Iterable<Header> headers) { this(topic, partition, (Long)null, key, value, headers); } public ProducerRecord(String topic, Integer partition, K key, V value) { this(topic, partition, (Long)null, key, value, (Iterable)null); } public ProducerRecord(String topic, K key, V value) { this(topic, (Integer)null, (Long)null, key, value, (Iterable)null); } public ProducerRecord(String topic, V value) { this(topic, (Integer)null, (Long)null, (Object)null, value, (Iterable)null); } public String topic() { return this.topic; } public Headers headers() { return this.headers; } public K key() { return this.key; } public V value() { return this.value; } public Long timestamp() { return this.timestamp; } public Integer partition() { return this.partition; } public String toString() { String headers = this.headers == null ? "null" : this.headers.toString(); String key = this.key == null ? "null" : this.key.toString(); String value = this.value == null ? "null" : this.value.toString(); String timestamp = this.timestamp == null ? "null" : this.timestamp.toString(); return "ProducerRecord(topic=" + this.topic + ", partition=" + this.partition + ", headers=" + headers + ", key=" + key + ", value=" + value + ", timestamp=" + timestamp + ")"; } public boolean equals(Object o) { if (this == o) { return true; } else if (!(o instanceof ProducerRecord)) { return false; } else { ProducerRecord<?, ?> that = (ProducerRecord)o; return Objects.equals(this.key, that.key) && Objects.equals(this.partition, that.partition) && Objects.equals(this.topic, that.topic) && Objects.equals(this.headers, that.headers) && Objects.equals(this.value, that.value) && Objects.equals(this.timestamp, that.timestamp); } } public int hashCode() { int result = this.topic != null ? this.topic.hashCode() : 0; result = 31 * result + (this.partition != null ? this.partition.hashCode() : 0); result = 31 * result + (this.headers != null ? this.headers.hashCode() : 0); result = 31 * result + (this.key != null ? this.key.hashCode() : 0); result = 31 * result + (this.value != null ? this.value.hashCode() : 0); result = 31 * result + (this.timestamp != null ? this.timestamp.hashCode() : 0); return result; } }
可以看到ProducerRecord在初始化时需要以下参数:
String topic;//分区名称,不可以为空
Integer partition;//当前记录需要写入的分区值,可以为空
Headers headers;//kafka头信息,可以为空
K key;//当前记录的key,可以为空
V value;//当前记录的实际value,不可以为空
Long timestamp;//指定生产者创建当前记录的时间戳,可以为空
在ProducerRecord的多个重构的构造函数中,参数最少的一个只需要传入topic名称和value即可。其他值都可为空。
public ProducerRecord(String topic, V value) { this(topic, (Integer)null, (Long)null, (Object)null, value, (Iterable)null); }
自定义序列化器示例
基于上述知识,我们可以通过实现KafkaSerializationSchema自定义序列化器,以下是一个最简单的自定义序列化器
package lenrnflink; import org.apache.flink.streaming.api.datastream.DataStream; import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment; import org.apache.flink.streaming.connectors.kafka.FlinkKafkaProducer; import org.apache.flink.streaming.connectors.kafka.KafkaSerializationSchema; import org.apache.kafka.clients.producer.ProducerRecord; import javax.annotation.Nullable; import java.util.Properties; public class Test { //实现KafkaSerializationSchema接口来自定义序列化器,传入的参数设定为String类型。 public static class MySerializationSchema implements KafkaSerializationSchema<String> { @Override public ProducerRecord<byte[], byte[]> serialize(String element, @Nullable Long timestamp) { //返回一条记录,指定topic为test,分区为0,key为null,value为传入对象转化而成的字节数组。 return new ProducerRecord<>("test",0,null,element.getBytes()); } } public static void main(String[] args) throws Exception { StreamExecutionEnvironment environment = StreamExecutionEnvironment.getExecutionEnvironment(); Properties properties = new Properties(); properties.setProperty("bootstrap.servers", "cdh1:9092,cdh2:9092,cdh3:9092,cdh4:9092,cdh5:9092"); //初始化Flink-Kafka生产者,并将自定义序列化器作为参数传递。 FlinkKafkaProducer<String> kafkaSink = new FlinkKafkaProducer<String>("test",new MySerializationSchema(),properties,FlinkKafkaProducer.Semantic.AT_LEAST_ONCE); DataStream<String> dataStream = environment.readTextFile("E:/test.txt","GB2312"); dataStream.addSink(kafkaSink); environment.execute("WordCount"); } }
以上只是一个最简单的自定义序列化器,用户在自己编写的时候,可以结合业务逻辑,根据传递过来的对象,在序列化器中指定其topic,partition,key和value等。
Flink中的Kafka分区器
源码解读
在Flink中,自定义Kafka分区器需要继承FlinkKafkaPartitioner抽象类,看一下源码:
@PublicEvolving public abstract class FlinkKafkaPartitioner<T> implements Serializable { private static final long serialVersionUID = -9086719227828020494L; //当前方法接收上游传过来的并行实例ID和并行实例总数。 //不是抽象方法,可以不用在子类实现。 public void open(int parallelInstanceId, int parallelInstances) { // overwrite this method if needed. } /** * @param record the record value * @param key serialized key of the record * @param value serialized value of the record * @param targetTopic target topic for the record * @param partitions found partitions for the target topic * @return the id of the target partition */ //当前方法为抽象方法,需要在子类中实现其方法体。指定其具体的分区。 public abstract int partition(T record, byte[] key, byte[] value, String targetTopic, int[] partitions); }
此类主要为open方法用于接收上游传过来的并行实例ID和总数。和partition抽象方法,进行指定分区的具体操作。
Flink中根据此实现了一个默认的分区器FlinkFixedPartitioner,看其源码:
@PublicEvolving public class FlinkFixedPartitioner<T> extends FlinkKafkaPartitioner<T> { private static final long serialVersionUID = -3785320239953858777L; private int parallelInstanceId; @Override public void open(int parallelInstanceId, int parallelInstances) { Preconditions.checkArgument(parallelInstanceId >= 0, "Id of this subtask cannot be negative."); Preconditions.checkArgument(parallelInstances > 0, "Number of subtasks must be larger than 0."); //接收上游传过来的并行实例ID,并将值传递给成员变量,用于分区操作。 this.parallelInstanceId = parallelInstanceId; } @Override public int partition(T record, byte[] key, byte[] value, String targetTopic, int[] partitions) { Preconditions.checkArgument( partitions != null && partitions.length > 0, "Partitions of the target topic is empty."); //通过Flink并行实例的id去和Kafka分区的数量取余来决定这个实例的数据写到哪个Kafka分区 return partitions[parallelInstanceId % partitions.length]; } @Override public boolean equals(Object o) { return this == o || o instanceof FlinkFixedPartitioner; } @Override public int hashCode() { return FlinkFixedPartitioner.class.hashCode(); } }
通过源码可以看出,当前分区器主要是通过Flink并行实例的id和Kafka分区的数量取余来决定这个实例的数据写到哪个Kafka分区,并且一个实例只写Kafka中的一个分区。
这样做的好处最大限度的利用了Flink和Kafka的可扩展性,提高数据处理效率。
自定义分区器示例
package lenrnflink; import org.apache.flink.api.common.serialization.SimpleStringSchema; import org.apache.flink.streaming.api.datastream.DataStream; import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment; import org.apache.flink.streaming.connectors.kafka.FlinkKafkaProducer; import org.apache.flink.streaming.connectors.kafka.KafkaSerializationSchema; import org.apache.flink.streaming.connectors.kafka.partitioner.FlinkKafkaPartitioner; import org.apache.kafka.clients.producer.ProducerRecord; import javax.annotation.Nullable; import java.util.Optional; import java.util.Properties; public class Test { //继承FlinkKafkaPartitioner自定义分区器 public static class KafkaPartitioner extends FlinkKafkaPartitioner<String> { @Override //当前分区器的逻辑为,如果传入的对象开头为1,则分区的值返回1,开头为2,则分区的值返回2,其他则返回0。 public int partition(String s, byte[] bytes, byte[] bytes1, String s2, int[] ints) { if(s.startsWith("1")){ return 1; } if(s.startsWith("2")){ return 2; } return 0; } } public static void main(String[] args) throws Exception { StreamExecutionEnvironment environment = StreamExecutionEnvironment.getExecutionEnvironment(); Properties properties = new Properties(); properties.setProperty("bootstrap.servers", "cdh1:9092,cdh2:9092,cdh3:9092,cdh4:9092,cdh5:9092"); //必须通过Optional容器添加自定义分区器。 Optional<FlinkKafkaPartitioner> ps = Optional.of(new KafkaPartitioner()); //初始化Flink-kafka生产者时,指定自定义分区器。 FlinkKafkaProducer kafkaSink = new FlinkKafkaProducer("test",new SimpleStringSchema(), properties, ps); DataStream<String> dataStream = environment.readTextFile("E:/test.txt","GB2312"); dataStream.addSink(kafkaSink); environment.execute("WordCount"); } }
以上为一个非常简单的自定义分区器,用户可根据业务需要制定一个属于自己的分区器。
结束语
以下为FlinkKafkaProducer源码
/** @deprecated */ @Deprecated public FlinkKafkaProducer(String brokerList, String topicId, SerializationSchema<IN> serializationSchema) { this(topicId, (KeyedSerializationSchema)(new KeyedSerializationSchemaWrapper(serializationSchema)), getPropertiesFromBrokerList(brokerList), (Optional)Optional.of(new FlinkFixedPartitioner())); } /** @deprecated */ @Deprecated public FlinkKafkaProducer(String topicId, SerializationSchema<IN> serializationSchema, Properties producerConfig) { this(topicId, (KeyedSerializationSchema)(new KeyedSerializationSchemaWrapper(serializationSchema)), producerConfig, (Optional)Optional.of(new FlinkFixedPartitioner())); } /** @deprecated */ @Deprecated public FlinkKafkaProducer(String topicId, SerializationSchema<IN> serializationSchema, Properties producerConfig, Optional<FlinkKafkaPartitioner<IN>> customPartitioner) { this(topicId, (KeyedSerializationSchema)(new KeyedSerializationSchemaWrapper(serializationSchema)), producerConfig, (Optional)customPartitioner); } /** @deprecated */ @Deprecated public FlinkKafkaProducer(String brokerList, String topicId, KeyedSerializationSchema<IN> serializationSchema) { this(topicId, serializationSchema, getPropertiesFromBrokerList(brokerList), Optional.of(new FlinkFixedPartitioner())); } /** @deprecated */ @Deprecated public FlinkKafkaProducer(String topicId, KeyedSerializationSchema<IN> serializationSchema, Properties producerConfig) { this(topicId, serializationSchema, producerConfig, Optional.of(new FlinkFixedPartitioner())); } /** @deprecated */ @Deprecated public FlinkKafkaProducer(String topicId, KeyedSerializationSchema<IN> serializationSchema, Properties producerConfig, FlinkKafkaProducer.Semantic semantic) { this(topicId, serializationSchema, producerConfig, Optional.of(new FlinkFixedPartitioner()), semantic, 5); } /** @deprecated */ @Deprecated public FlinkKafkaProducer(String defaultTopicId, KeyedSerializationSchema<IN> serializationSchema, Properties producerConfig, Optional<FlinkKafkaPartitioner<IN>> customPartitioner) { this(defaultTopicId, serializationSchema, producerConfig, customPartitioner, FlinkKafkaProducer.Semantic.AT_LEAST_ONCE, 5); } /** @deprecated */ @Deprecated public FlinkKafkaProducer(String defaultTopicId, KeyedSerializationSchema<IN> serializationSchema, Properties producerConfig, Optional<FlinkKafkaPartitioner<IN>> customPartitioner, FlinkKafkaProducer.Semantic semantic, int kafkaProducersPoolSize) { this(defaultTopicId, serializationSchema, (FlinkKafkaPartitioner)customPartitioner.orElse((Object)null), (KafkaSerializationSchema)null, producerConfig, semantic, kafkaProducersPoolSize); } public FlinkKafkaProducer(String defaultTopic, KafkaSerializationSchema<IN> serializationSchema, Properties producerConfig, FlinkKafkaProducer.Semantic semantic) { this(defaultTopic, serializationSchema, producerConfig, semantic, 5); } public FlinkKafkaProducer(String defaultTopic, KafkaSerializationSchema<IN> serializationSchema, Properties producerConfig, FlinkKafkaProducer.Semantic semantic, int kafkaProducersPoolSize) { this(defaultTopic, (KeyedSerializationSchema)null, (FlinkKafkaPartitioner)null, serializationSchema, producerConfig, semantic, kafkaProducersPoolSize); }
在阅读Flink中的Kafka生产者源码FlinkKafkaProducer时发现其多个构造函数,凡是参数中包含FlinkKafkaProducer的都被标记为了deprecated,说明官方已经不推荐使用自定义分区器来进行数据的分区操作。
只有参数包含KafkaSerializationSchema的两个构造函数是正常的,说明现在官方推荐使用KafkaSerializationSchema接口来进行序列化的操作。
并且阅读源码的过程中可以发现,KafkaSerializationSchema中也有对数据的分区操作。只需要结合KafkaContextAware接口即可实现获取Flink并行实例ID和数量的功能。