spring-cloud-stream-binder-kafka属性配置

序

本文简单介绍下spring-cloud-stream-binder-kafka的一些属性配置。

maven

       <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-stream-binder-kafka</artifactId>
            <version>1.0.3.RELEASE</version>
        </dependency>

stream属性

spring-cloud-stream-1.0.3.RELEASE-sources.jar!/org/springframework/cloud/stream/config/ChannelBindingServiceProperties.java

spring:
  cloud:
     stream:
       instanceIndex: 0 ##支持环境变量INSTANCE_INDEX
                        ## The instance index of the application: a number from 0 to instanceCount-1. Used for partitioning and with Kafka
       instanceCount: 1 ## The number of deployed instances of an application. Must be set for partitioning and if using Kafka.
                        ## used to partition data across different consumers.

Topic在逻辑上可以被认为是一个queue。每条消费都必须指定它的topic，可以简单理解为必须指明把这条消息放进哪个queue里。为了使得 Kafka的吞吐率可以水平扩展，物理上把topic分成一个或多个partition，每个partition在物理上对应一个文件夹，该文件夹下存储 这个partition的所有消息和索引文件。partiton命名规则为topic名称+有序序号，第一个partiton序号从0开始，序号最大值为partitions数量减1。 同一个partition内的消息只能被同一个组中的一个consumer消费。 当消费者数量多于partition的数量时，多余的消费者空闲。 消费者少于和等于partition的数量时，会出现多个partition对应一个消费者的情况，个别消费者消费量会比其他的多。

instanceCount主要是consumer用的，一般小于或等于topic的partition数量，主要用作消费者的消费分区用。

bingdings属性

spring-cloud-stream-1.0.3.RELEASE-sources.jar!/org/springframework/cloud/stream/config/BindingProperties.java

spring:
  cloud:
     stream:
        bindings:
          output:
            destination: event-demo
            content-type: text/plain
            #group: test  ##consumer属性
            #producer:
            #consumer:

producer

spring-cloud-stream-1.0.3.RELEASE-sources.jar!/org/springframework/cloud/stream/binder/ProducerProperties.java

spring:
  cloud:
     stream:
        bindings:
          output:
            destination: event-demo
            content-type: text/plain
            producer:
              partitionCount: 1
              headerMode
              partitionKeyExtractorClass: org.springframework.cloud.stream.partitioning.CustomPartitionKeyExtractorClass
              partitionSelectorClass: org.springframework.cloud.stream.partitioning.CustomPartitionSelectorClass
              headerMode: raw

• kafka producer扩展属性 spring-cloud-stream-binder-kafka-1.0.3.RELEASE-sources.jar!/org/springframework/cloud/stream/binder/kafka/KafkaProducerProperties.java spring: cloud: stream: bindings: output: destination: event-demo content-type: text/plain producer: bufferSize: 16384 maxRequestSize: 1048576 sync: true batchTimeout: 0

consumer

spring-cloud-stream-1.0.3.RELEASE-sources.jar!/org/springframework/cloud/stream/binder/ConsumerProperties.java

spring:
  cloud:
     stream:
        bindings:
          input:
            destination: event-demo
            content-type: text/plain
            consumer:
              concurrency: 1 ## The concurrency of the inbound consumer.
              partitioned: false ## Whether the consumer receives data from a partitioned producer.Default: false.
              headerMode: raw

• kafka consumer扩展属性 spring-cloud-stream-binder-kafka-1.0.3.RELEASE-sources.jar!/org/springframework/cloud/stream/binder/kafka/KafkaConsumerProperties.java spring: cloud: stream: bindings: input: destination: event-demo content-type: text/plain consumer: autoCommitOffset: false resetOffsets: true startOffset: earliest enableDlq: false recoveryInterval: 5000

原生api

ConsumerConfig consumerConfig =  new kafka.consumer.ConsumerConfig(props);
        ConsumerConnector consumerConnector = kafka.consumer.Consumer.createJavaConsumerConnector(consumerConfig);
        Map<String, Integer> topicCountMap = new HashMap<String, Integer>();
        topicCountMap.put(topic, consumerCount);
        Map<String, List<KafkaStream<byte[], byte[]>>> consumerMap = consumerConnector
                .createMessageStreams(topicCountMap);

这里头topicCountMap告诉Kafka我们在Consumer中将用多少个线程来消费该topic。topicCountMap的key是topic name，value针对该topic是线程的数量。

小结

整体的话，spring cloud stream自己抽象了一部分，但是有个硬伤就是spring.cloud.stream.instanceIndex这个不大友好，这样就造成服务的实例是有状态的了，在基于docker部署起来比较麻烦，还不如直接原生api。如果partition不多，或者每个consumer性能强悍的话，那么至少部署两个，配置起来也还可以接受。

doc

• spring-cloud-stream-binder-kafka-docs
• spring-cloud-stream-docs
• SpringCloudStream 构建消息驱动的微服务框架
• kafka中partition和消费者对应关系
• kafka学习(四)-Topic & Partition