kafka7 探索生产者同步or异步发送消息
1.生产者:在发送完消息后,收到回执确认。
主要是在SimpleProducer.java中修改了发送消息的2行代码,用到了回调函数,修改如下:
//发送消息 ProducerRecord<String, String> rec = new ProducerRecord<String, String>("test-topic","hello world from win7"); producer.send(rec,new Callback() { public void onCompletion(RecordMetadata metadata,Exception exception) { System.out.println("ack!!!"); } }); //在发送消息后,收到回执确认。
完整代码如下:
1 package cn.test.mykafka; 2 3 import java.util.Properties; 4 5 import org.apache.kafka.clients.producer.Callback; 6 import org.apache.kafka.clients.producer.KafkaProducer; 7 import org.apache.kafka.clients.producer.Producer; 8 import org.apache.kafka.clients.producer.ProducerRecord; 9 import org.apache.kafka.clients.producer.RecordMetadata; 10 11 /** 12 * 简单生产者:在发送完消息后,收到回执确认。 13 * 14 */ 15 16 public class SimpleProducer2 { 17 18 public static void main(String[] args) { 19 20 //创建配置信息 21 Properties props = new Properties(); 22 props.put("bootstrap.servers", "192.168.42.133:9092"); //指定broker的节点和端口 23 props.put("acks", "all"); 24 props.put("retries", 0); 25 props.put("batch.size", 16384); 26 props.put("linger.ms", 1); 27 props.put("buffer.memory", 33554432); 28 props.put("key.serializer", "org.apache.kafka.common.serialization.StringSerializer"); 29 props.put("value.serializer", "org.apache.kafka.common.serialization.StringSerializer"); 30 31 //创建一个生产者 32 Producer<String, String> producer = new KafkaProducer<>(props); 33 34 //发送消息 35 ProducerRecord<String, String> rec = new ProducerRecord<String, String>("test-topic","hello world from win7"); 36 producer.send(rec,new Callback() { 37 public void onCompletion(RecordMetadata metadata,Exception exception) { 38 System.out.println("ack!!!"); 39 } 40 }); //在发送消息后,收到回执确认。 41 42 //for (int i = 0; i < 10; i++) 43 // producer.send(new ProducerRecord<String, String>("test-topic", Integer.toString(i), Integer.toString(i))); //topic,key(非必填),value 44 45 System.out.println("over"); 46 producer.close(); 47 } 48 }
2.比较同步和异步生产者消息发送速度。
完整代码如下:
1 package cn.test.mykafka; 2 3 import java.util.Properties; 4 5 import org.apache.kafka.clients.producer.Callback; 6 import org.apache.kafka.clients.producer.KafkaProducer; 7 import org.apache.kafka.clients.producer.Producer; 8 import org.apache.kafka.clients.producer.ProducerRecord; 9 import org.apache.kafka.clients.producer.RecordMetadata; 10 11 /** 12 * 比较同步和异步生产者消息发送速度,参数未生效,失败 13 * 14 */ 15 16 public class SimpleProducer3 { 17 18 static long starttime; 19 public static void main(String[] args) { 20 21 //创建配置信息 22 Properties props = new Properties(); 23 props.put("bootstrap.servers", "192.168.42.133:9092"); //指定broker的节点和端口 24 props.put("acks", "all"); 25 props.put("retries", 0); 26 props.put("batch.size", 16384); 27 props.put("linger.ms", 1); 28 props.put("buffer.memory", 33554432); 29 props.put("key.serializer", "org.apache.kafka.common.serialization.StringSerializer"); 30 props.put("value.serializer", "org.apache.kafka.common.serialization.StringSerializer"); 31 props.put("producer.type", "async"); //sync为同步,async为异步,此配置未生效 32 33 //创建一个生产者 34 Producer<String, String> producer = new KafkaProducer<>(props); 35 36 StringBuilder builder = new StringBuilder(); 37 for( int i = 0 ; i < 100000 ; i++) { 38 builder.append(" " + i + ","); 39 } 40 41 //发送消息 42 ProducerRecord<String, String> rec = new ProducerRecord<String, String>("test-topic",builder.toString()); 43 producer.send(rec,new Callback() { 44 public void onCompletion(RecordMetadata metadata,Exception exception) { 45 System.out.println("receive ack : "+ (System.currentTimeMillis()-starttime) + "ms"); 46 } 47 }); //在发送消息后,收到回执确认 48 49 starttime = System.currentTimeMillis(); 50 System.out.println("over"); 51 producer.close(); 52 } 53 }
kafka同步生产者:这个生产者写一条消息的时候,它就立马发送到某个分区去。follower还需要从leader拉取消息到本地,follower再向leader发送确认,leader再向客户端发送确认。由于这一套流程之后,客户端才能得到确认,所以很慢。
kafka异步生产者:这个生产者写一条消息的时候,先是写到某个缓冲区,这个缓冲区里的数据还没写到broker集群里的某个分区的时候,它就返回到client去了。虽然效率快,但是不能保证消息一定被发送出去了。
客户端向topic发送数据分为两种方式:
producer.type=sync 同步模式
producer.type=async 异步模式
执行以上代码(通过控制producer.type参数取值sync/async)时,输出警告:
WARN org.apache.kafka.clients.producer.ProducerConfig - The configuration 'producer.type' was supplied but isn't a known config.
这说明producer.type参数配置根本没生效,后来我在官方文档中也没有找到这个参数,估计在kafka 2.0.0版本中此参数已经没有了。
于是乎,我在网上找了另一段代码(参考博客),修改后如下:
1 package cn.test.mykafka; 2 3 import java.util.Properties; 4 import java.util.concurrent.Future; 5 6 import org.apache.kafka.clients.producer.Callback; 7 import org.apache.kafka.clients.producer.KafkaProducer; 8 import org.apache.kafka.clients.producer.Producer; 9 import org.apache.kafka.clients.producer.ProducerRecord; 10 import org.apache.kafka.clients.producer.RecordMetadata; 11 12 /** 13 * 比较同步和异步生产者消息发送速度,有2个问题: 1.我不缺定starttime的取值位置是否正确? 2.时间差不多,没得出啥结论 14 * 15 */ 16 17 public class SimpleProducer4 { 18 19 static long starttime; 20 21 StringBuilder builder = new StringBuilder(); 22 23 public void initBuilder() { 24 for (int i = 0; i < 100000; i++) { 25 builder.append(" " + i + ","); 26 } 27 } 28 29 private Properties kafkaProps = new Properties(); 30 31 /** 32 * 初始化一些配置信息 33 */ 34 public void initProperty() { 35 kafkaProps.put("bootstrap.servers", "192.168.42.133:9092"); // 指定broker的节点和端口 36 kafkaProps.put("acks", "all"); 37 kafkaProps.put("retries", 0); 38 kafkaProps.put("batch.size", 16384); 39 kafkaProps.put("linger.ms", 1); 40 kafkaProps.put("buffer.memory", 33554432); 41 kafkaProps.put("key.serializer", "org.apache.kafka.common.serialization.StringSerializer"); 42 kafkaProps.put("value.serializer", "org.apache.kafka.common.serialization.StringSerializer"); 43 } 44 45 /** 46 * 加载配置信息,生成一个生产者实例 47 * 48 * @param props 49 * @return 50 */ 51 public Producer<String, String> getProducer(Properties props) { 52 if (props == null || props.size() == 0) 53 throw new IllegalArgumentException(); 54 return new KafkaProducer<>(props); 55 } 56 57 /** 58 * 同步发送消息 59 * 60 * @param producer 61 * @throws Exception 62 */ 63 public void syncSend(Producer<String, String> producer) throws Exception { 64 65 ProducerRecord<String, String> record = new ProducerRecord<String, String>("test-topic", 66 builder.toString() + "this is a sygn record"); 67 68 // 同步发送消息,消息发送成功后,服务端会返回给一个RecordMetadata对象 69 Future<RecordMetadata> future = producer.send(record); 70 starttime = System.currentTimeMillis(); 71 RecordMetadata metadata = future.get(); 72 73 System.out.println("offset:" + metadata.offset() + "\npartition:" + metadata.partition() + "\ntopic:" 74 + metadata.topic() + "\nserializedKeySize:" + metadata.serializedKeySize() + "\nserializedValueSize:" 75 + metadata.serializedValueSize() + "\nreceive sygn ack : " + (System.currentTimeMillis() - starttime) 76 + "ms" + "\n"); 77 78 producer.close(); 79 } 80 81 /** 82 * 异步发送消息 83 * 84 * @param producer 85 */ 86 public void asyncSend(Producer<String, String> producer) { 87 88 ProducerRecord<String, String> record = new ProducerRecord<String, String>("test-topic", 89 builder.toString() + "this is a asygn record"); 90 91 producer.send(record, new Callback() { 92 public void onCompletion(RecordMetadata metadata, Exception e) { 93 System.out.println("offset:" + metadata.offset() + "\npartition:" + metadata.partition() + "\ntopic:" 94 + metadata.topic() + "\nserializedKeySize:" + metadata.serializedKeySize() 95 + "\nserializedValueSize:" + metadata.serializedValueSize() + "\n"); 96 if (e == null) { 97 System.out.println("\nreceive asygn ack : " + (System.currentTimeMillis() - starttime) + "ms"); 98 } 99 } 100 }); 101 starttime = System.currentTimeMillis(); 102 producer.close(); 103 } 104 105 public void start() throws Exception { 106 initBuilder(); 107 initProperty(); 108 // syncSend(getProducer(kafkaProps)); 109 asyncSend(getProducer(kafkaProps)); 110 111 } 112 113 public static void main(String[] args) throws Exception { 114 115 SimpleProducer4 myProducer = new SimpleProducer4(); 116 myProducer.start(); 117 } 118 }
执行之后,发现同步和异执行时间差不多,没能证明异步比同步快。推测原因有2个:
1.消息太短,存在偶然性,看不出时间差;
2.我不确定这段代码是否正确?同步代码是否正确?异步代码是否正确?开始时间取值位置是否正确?
好吧,心好累。关于生产者的同步异步问题就先到这里吧,虽然结果不尽人意,但是过程中也学到了很多,以后有时间再继续吧。
PS:以上纯粹是我的探索测试,如果有不对的地方,欢迎留言指正,不胜感激。