Kafka：消费-转换-生产模式的事务

爱被打了一巴掌 2024-03-24 13:28 71阅读 0赞

import org.apache.kafka.clients.consumer.*;
    import org.apache.kafka.clients.producer.KafkaProducer;
    import org.apache.kafka.clients.producer.ProducerConfig;
    import org.apache.kafka.clients.producer.ProducerRecord;
    import org.apache.kafka.common.TopicPartition;
    import org.apache.kafka.common.serialization.StringDeserializer;
    import org.apache.kafka.common.serialization.StringSerializer;
    
    public class MyTransactional {
    
        public static KafkaProducer<String, String> getProducer() {
            Map<String, Object> configs = new HashMap<>();
            configs.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, "192.168.25.129:9092");
            configs.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
            configs.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
    
            // 设置client.id
            configs.put(ProducerConfig.CLIENT_ID_CONFIG, "tx_producer_01");
    
            // 设置事务id
            configs.put(ProducerConfig.TRANSACTIONAL_ID_CONFIG, "tx_id_01");
    
            // 需要所有的ISR副本确认
            configs.put(ProducerConfig.ACKS_CONFIG, "all");
    
            // 启用幂等性
            configs.put(ProducerConfig.ENABLE_IDEMPOTENCE_CONFIG, true);
    
            KafkaProducer<String, String> producer = new KafkaProducer<String, String>(configs);
    
            return producer;
        }
    
        public static KafkaConsumer<String, String> getConsumer(String consumerGroupId) {
            Map<String, Object> configs = new HashMap<>();
            configs.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, "192.168.25.129:9092");
            configs.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class);
            configs.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class);
    
            // 设置消费组ID
            configs.put(ConsumerConfig.GROUP_ID_CONFIG, "consumer_grp_01");
    
            // 不启用消费者偏移量的自动确认，也不要手动确认
            configs.put(ConsumerConfig.ENABLE_AUTO_COMMIT_CONFIG, false);
    
            configs.put(ConsumerConfig.CLIENT_ID_CONFIG, "consumer_client_01");
    
            configs.put(ConsumerConfig.AUTO_OFFSET_RESET_CONFIG, "earliest");
    
            // 只读取已提交的消息
            // 消费-转换-生产 模式中生产者下游的消费者开启read_committed，默认为read_uncommitted
    //        configs.put(ConsumerConfig.ISOLATION_LEVEL_CONFIG, "read_committed");
    
            KafkaConsumer<String, String> consumer = new KafkaConsumer<String, String>(configs);
    
            return consumer;
        }
    
    
        public static void main(String[] args) {
    
            String consumerGroupId = "consumer_grp_id_01";
    
            KafkaProducer<String, String> producer = getProducer();
            KafkaConsumer<String, String> consumer = getConsumer(consumerGroupId);
    
            // 事务的初始化
            producer.initTransactions();
            //订阅主题
            consumer.subscribe(Collections.singleton("tp_tx_01"));
    
            final ConsumerRecords<String, String> records = consumer.poll(1_000);
    
            // 开启事务
            producer.beginTransaction();
            try {
                Map<TopicPartition, OffsetAndMetadata> offsets = new HashMap<>();
    
                for (ConsumerRecord<String, String> record : records) {
    
                    System.out.println(record);
    
                    producer.send(new ProducerRecord<String, String>("tp_tx_out_01", record.key(), record.value()));
    
                    offsets.put(
                            new TopicPartition(record.topic(), record.partition()),
                            new OffsetAndMetadata(record.offset() + 1)); // 偏移量表示下一条要消费的消息
                }
    
                // 将该消息的偏移量提交作为事务的一部分，随事务提交和回滚（不提交消费偏移量）
                producer.sendOffsetsToTransaction(offsets, consumerGroupId);
    
    //            int i = 1 / 0;
    
                // 提交事务
                producer.commitTransaction();
    
            } catch (Exception e) {
                e.printStackTrace();
                // 回滚事务
                producer.abortTransaction();
            } finally {
                // 关闭资源
                producer.close();
                consumer.close();
            }
    
    
        }
    }