kafka源码对 java 和 gradle 版本有适配要求，否则导入源码会编译不通过，笔者引用各版本如下。
| Kafka源码 | 3.0 |
|---|---|
| java版本 | 11 |
| grade版本 | 7.4.1 |
上图对 kafka 生产者初始化的整体流程进行了简单的梳理，接下来我们从代码层面对整体流程进行一个复原。
源码导入idea后都会看到一个example工程包，这个包里面有生产者和消费者的例子，可以作为源码学习的切入口。
在初始化生产者之前会设置一些配置，包括kafka地址、key/value序列化器等，当然上面的配置在我们实际使用时是远远不够的（后续在分析过程中会整理一些kafka调优的参数），通过层层构造函数，会进入到 KafkaProducer 核心构造方法，完整代码如下：
KafkaProducer(ProducerConfig config,
Serializer keySerializer,
Serializer valueSerializer,
Producermetadata metadata,
KafkaClient kafkaClient,
ProducerInterceptors interceptors,
Time time) {
try {
this.producerConfig = config;
this.time = time;
String transactionalId = config.getString(ProducerConfig.TRANSACTIONAL_ID_CONFIG);// 获取事务id
this.clientId = config.getString(ProducerConfig.CLIENT_ID_CONFIG); // 获取客户端id 如果没有设置系统会每个都默认会生成一个client.id，producer-自增长的数字，producer-1，producer-2等
LogContext logContext;
if (transactionalId == null)
logContext = new LogContext(String.format("[Producer clientId=%s] ", clientId));
else
logContext = new LogContext(String.format("[Producer clientId=%s, transactionalId=%s] ", clientId, transactionalId));
log = logContext.logger(KafkaProducer.class);
log.trace("Starting the Kafka producer");
Map metricTags = Collections.singletonMap("client-id", clientId); //创建一个不可变的Map集合
MetricConfig metricConfig = new MetricConfig()
.samples(config.getInt(ProducerConfig.METRICS_NUM_SAMPLES_CONFIG))
.timeWindow(config.getLong(ProducerConfig.METRICS_SAMPLE_WINDOW_MS_CONFIG), TimeUnit.MILLISECONDS)
.recordLevel(Sensor.RecordingLevel.forName(config.getString(ProducerConfig.METRICS_RECORDING_LEVEL_CONFIG)))
.tags(metricTags);
List reporters = config.getConfiguredInstances(ProducerConfig.METRIC_REPORTER_CLASSES_CONFIG,
MetricsReporter.class,
Collections.singletonMap(ProducerConfig.CLIENT_ID_CONFIG, clientId));
JmxReporter jmxReporter = new JmxReporter();
jmxReporter.configure(config.originals(Collections.singletonMap(ProducerConfig.CLIENT_ID_CONFIG, clientId)));
reporters.add(jmxReporter);
MetricsContext metricsContext = new KafkaMetricsContext(JMX_PREFIX,
config.originalsWithPrefix(CommonClientConfigs.METRICS_CONTEXT_PREFIX));
this.metrics = new Metrics(metricConfig, reporters, time, metricsContext);
this.partitioner = config.getConfiguredInstance(
ProducerConfig.PARTITIONER_CLASS_CONFIG,
Partitioner.class,
Collections.singletonMap(ProducerConfig.CLIENT_ID_CONFIG, clientId));
long retryBackoffMs = config.getLong(ProducerConfig.RETRY_BACKOFF_MS_CONFIG);
if (keySerializer == null) {
this.keySerializer = config.getConfiguredInstance(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, Serializer.class);
this.keySerializer.configure(config.originals(Collections.singletonMap(ProducerConfig.CLIENT_ID_CONFIG, clientId)), true);
} else {
config.ignore(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG);
this.keySerializer = keySerializer;
}
if (valueSerializer == null) {
this.valueSerializer = config.getConfiguredInstance(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, Serializer.class);
this.valueSerializer.configure(config.originals(Collections.singletonMap(ProducerConfig.CLIENT_ID_CONFIG, clientId)), false);
} else {
config.ignore(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG);
this.valueSerializer = valueSerializer;
}
List> interceptorList = (List) config.getConfiguredInstances(
ProducerConfig.INTERCEPTOR_CLASSES_CONFIG,
ProducerInterceptor.class,
Collections.singletonMap(ProducerConfig.CLIENT_ID_CONFIG, clientId));
if (interceptors != null)
this.interceptors = interceptors;
else
this.interceptors = new ProducerInterceptors<>(interceptorList);
// 抽象了一个接收元数据更新集群资源的监听器集合。
ClusterResourceListeners clusterResourceListeners = configureClusterResourceListeners(keySerializer,
valueSerializer, interceptorList, reporters);
this.maxRequestSize = config.getInt(ProducerConfig.MAX_REQUEST_SIZE_CONFIG);
this.totalMemorySize = config.getLong(ProducerConfig.BUFFER_MEMORY_CONFIG);
this.compressionType = CompressionType.forName(config.getString(ProducerConfig.COMPRESSION_TYPE_CONFIG));
this.maxBlockTimeMs = config.getLong(ProducerConfig.MAX_BLOCK_MS_CONFIG);
int deliveryTimeoutMs = configureDeliveryTimeout(config, log);
this.apiVersions = new ApiVersions();
this.transactionManager = configureTransactionState(config, logContext);
this.accumulator = new RecordAccumulator(logContext,
config.getInt(ProducerConfig.BATCH_SIZE_CONFIG), // 批次大小 默认16k
this.compressionType, // 压缩方式，默认是none
lingerMs(config),
retryBackoffMs,
deliveryTimeoutMs,
metrics,
PRODUCER_METRIC_GROUP_NAME,
time,
apiVersions,
transactionManager,
new BufferPool(this.totalMemorySize, // 缓冲区对象 默认是32m
config.getInt(ProducerConfig.BATCH_SIZE_CONFIG),
metrics, time, PRODUCER_METRIC_GROUP_NAME));
// 解析连接上kafka集群地址
List addresses = ClientUtils.parseAndValidateAddresses(
config.getList(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG),
config.getString(ProducerConfig.CLIENT_DNS_LOOKUP_CONFIG));// client.dns.lookup 控制客户端如何使用DNS查找
if (metadata != null) {
this.metadata = metadata;
} else {
this.metadata = new Producermetadata(retryBackoffMs,
config.getLong(ProducerConfig.metaDATA_MAX_AGE_CONFIG),
config.getLong(ProducerConfig.metaDATA_MAX_IDLE_CONFIG),
logContext,
clusterResourceListeners,
Time.SYSTEM);
this.metadata.bootstrap(addresses);
}
this.errors = this.metrics.sensor("errors");
this.sender = newSender(logContext, kafkaClient, this.metadata); // 本身是一个线程
String ioThreadName = NETWORK_THREAD_PREFIX + " | " + clientId; // 线程名称
// 把sender线程放到后台
this.ioThread = new KafkaThread(ioThreadName, this.sender, true);
// 启动sender线程 调用线程run方法
this.ioThread.start();
config.logUnused();
AppInfoParser.registerAppInfo(JMX_PREFIX, clientId, metrics, time.milliseconds());
log.debug("Kafka producer started");
} catch (Throwable t) {
// call close methods if internal objects are already constructed this is to prevent resource leak. see KAFKA-2121
close(Duration.ofMillis(0), true);
// now propagate the exception
throw new KafkaException("Failed to construct kafka producer", t);
}
}
分区器 Partitioner
Kafka发送每条消息都会有一个路由操作，其实就是被分配到哪个分区里去。我们可以通过指定生产者partitioner.class 参数实现数据自定义分区，系统默认分区器：DefaultPartitioner。
对于 producer 而言，interceptor 使得用户在消息发送前以及 produce r回调逻辑前有机会对消息做一些定制化需求，比如修改消息等。
kafka发送消息为了减少网络请求、提高吞吐，并不是直接将消息从客户端通过网络发送给服务器端，而是先将消息存储在客户端的记录收集器（缓冲区）中，当队列满了batch.size或者发送时间linger.ms已到的时候才会去发送，这个记录收集器就是RecordAccumulator。
当一条消息要写入broker，需要先知道这条数据要写入哪个分区及在哪个broker上，metaData是用来从broker集群去拉取元数据的Topics（Topic -> Partitions（Leader+Followers，ISR））
Kafka发消息默认是异步的，主线程生产消息，放在我们上面说的记录收集器（RecordAccumulator）里，另一个线程 Sender 拉取消息发送到Broker。
在 Sender 线程初始化之前会通过 NetworkClient 组件来构建网络传输桥梁。
关注公众号 ，专注于java大数据领域离线、实时技术干货定期分享！如有问题随时沟通交流！ www.lllpan.top
