topic: "topic_query_p3r1" 分配了三个partition分区
实现顺序性原理：
设置相同的key会把消息投递到同一个分区的topic中，再由一个消费者来消费该分区topic。
投递顺序消息
同一组行为设置相同的key，会把这组数据投递到同一分区topic中。
/**
* 投递顺序性消息，根据用户id做取模推送到不同分区的topic中
* 相同的key推送到同一分区中
*/
@RequestMapping("/kafka2")
public String testKafka2() {
for (int userId = 0; userId < 300; userId++) {
kafkaTemplate.send("topic_query_p3r1", userId + "", "insert" + userId);
kafkaTemplate.send("topic_query_p3r1", userId + "", "update" + userId);
kafkaTemplate.send("topic_query_p3r1", userId + "", "delete" + userId);
}
return null;
}
消费顺序消息
方式1 - 直接进行消费
因为投递的相同行为的消息是有序的，所以直接消费也不会有问题。
/**
* 消费topic_query_p3r1主题，ConsumerGroupId1消费组
*/
@KafkaListener(topics = "topic_query_p3r1", groupId = "ConsumerGroupId1")
public void p3r2ConsumerGroupId0(ConsumerRecord<?, ?> consumer) throws InterruptedException {
System.out.println("消费者A topic名称:" + consumer.topic() +
", key:" + consumer.key() +
", value:" + consumer.value() +
", 分区位置:" + consumer.partition() +
", 下标" + consumer.offset()+" "+Thread.currentThread().getId());
Thread.sleep(10);
}
方式2.1 - 一个消费者来指定具体分区进行消费
指定具体分区来进行消费。
/**
* 消费者，解决消息顺序性
* 注解参数：partitions=0表示：只消费该主题中0分区的数据。
*/
@KafkaListener(topicPartitions = {@TopicPartition(topic = "topic_query_p3r1", partitions = {"0"})}, groupId = "ConsumerGroupId1")
public void receive(ConsumerRecord<?, ?> consumer) {
System.out.println("消费者C topic名称:" + consumer.topic() +
",key:" + consumer.key() + "," +
",value:" + consumer.value() + "," +
"分区位置:" + consumer.partition() +
", 下标" + consumer.offset());
}
方式2.2 - 多个消费者来指定不同分区进行消费。
写多个消费者方法来分别指向不同分区，提高消费速度，但是此方法不灵活。
/**
* 消费0分区的topic_query_p3r1主题消费者，ConsumerGroupId1消费组
*/
@KafkaListener(topicPartitions = {@TopicPartition(topic = "topic_query_p3r1", partitions = {"0"})}, groupId = "ConsumerGroupId1")
public void p3r2ConsumerGroupId0(ConsumerRecord<?, ?> consumer) throws InterruptedException {
System.out.println("消费者A topic名称:" + consumer.topic() +
", key:" + consumer.key() +
", value:" + consumer.value() +
", 分区位置:" + consumer.partition() +
", 下标" + consumer.offset()+" "+Thread.currentThread().getId());
Thread.sleep(10);
}
/**
* 消费1分区的topic_query_p3r1主题消费者，ConsumerGroupId1消费组
*/
@KafkaListener(topicPartitions = {@TopicPartition(topic = "topic_query_p3r1", partitions = {"1"})}, groupId = "ConsumerGroupId1")
public void p3r2ConsumerGroupId1(ConsumerRecord<?, ?> consumer) throws InterruptedException {
System.out.println("消费者A topic名称:" + consumer.topic() +
", key:" + consumer.key() +
", value:" + consumer.value() +
", 分区位置:" + consumer.partition() +
", 下标" + consumer.offset()+" "+Thread.currentThread().getId());
Thread.sleep(10);
}
/**
* 消费2分区的topic_query_p3r1主题消费者，ConsumerGroupId1消费组
*/
@KafkaListener(topicPartitions = {@TopicPartition(topic = "topic_query_p3r1", partitions = {"2"})}, groupId = "ConsumerGroupId1")
public void p3r2ConsumerGroupId2(ConsumerRecord<?, ?> consumer) throws InterruptedException {
System.out.println("消费者A topic名称:" + consumer.topic() +
", key:" + consumer.key() +
", value:" + consumer.value() +
", 分区位置:" + consumer.partition() +
", 下标" + consumer.offset()+" "+Thread.currentThread().getId());
Thread.sleep(10);
}
多线程顺序消费
// 使用两个内存队列
final int queueLingth = 2;
// 创建两个内存队列
Queue<Map> queueA = new ConcurrentLinkedQueue<>();
Queue<Map> queueB = new ConcurrentLinkedQueue<>();
/**
* 投递顺序性消息，根据用户id做取模推送到不同分区的topic中
* 相同的key推送到相同的分区中
*/
@RequestMapping("/kafka2")
public String testKafka2() {
for (int userId = 0; userId < 300; userId++) {
kafkaTemplate.send("topic_query_p3r1", userId + "", "insert" + userId);
kafkaTemplate.send("topic_query_p3r1", userId + "", "update" + userId);
kafkaTemplate.send("topic_query_p3r1", userId + "", "delete" + userId);
}
return null;
}
/**
* 主题消费者-把相同行为的数据放到同一内存队列中
*/
@KafkaListener(topics = "topic_query_p3r1", groupId = "ConsumerGroupId1")
public void p3r2ConsumerGroupId0(ConsumerRecord<?, ?> consumer){
// 1.封装消息参数
Map param = new HashMap();
param.put("topic", consumer.topic());
param.put("key", consumer.key());
param.put("value", consumer.value());
param.put("p", consumer.partition());
// 2.把相同行为(key)数据添加到同一内存队列中
int queueHash = consumer.key().hashCode() % queueLingth;
if (queueHash == 0) {
queueA.add(param);
}
if (queueHash == 1) {
queueB.add(param);
}
}
// 开启两个线程消费内存队列中的消息
@Override
public void run(ApplicationArguments args) throws Exception {
new Thread() {
@Override
public void run() {
while (true) {
if (queueA.size() > 0) {
Map poll = queueA.poll();
System.out.println("Thrend-Id: "+ Thread.currentThread().getId() +
" topic：" + poll.get("topic") +
" key:" + poll.get("key") +
" value:" + poll.get("value") +
" partition:" + poll.get("p"));
try {
Thread.sleep(10);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
}
}.start();
new Thread() {
@Override
public void run() {
while (true) {
if (queueB.size() > 0) {
Map poll = queueB.poll();
System.out.println("Thrend-Id: "+ Thread.currentThread().getId() + " topic：" + poll.get("topic") + " key:" + poll.get("key") + " value:" + poll.get("value") + " partition:" + poll.get("p"));
try {
Thread.sleep(10);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
}
}.start();
}
打印：insert、update、delete都是有序的。相同行为都在同一线程下执行。
