导致kafka的重复消费问题原因在于,已经消费了数据,但是offset没来得及提交(比如Kafka没有或者不知道该数据已经被消费)。 总结以下场景导致Kakfa重复消费:
例如:
try {
consumer.unsubscribe();
} catch (Exception e) {
}
try {
consumer.close();
} catch (Exception e) {
}
上面代码会导致部分offset没提交,下次启动时会重复消费。
解决方法:设置offset自动提交为false
整合了Spring配置的修改如下配置 spring配置:
spring.kafka.consumer.enable-auto-commit=false
spring.kafka.consumer.auto-offset-reset=latest
整合了API方式的修改enable.auto.commit为false API配置:
Properties props = new Properties();
props.put("bootstrap.servers", "localhost:9092");
props.put("group.id", "test");
props.put("enable.auto.commit", "false");
一旦设置了 enable.auto.commit 为 true,Kafka 会保证在开始调用 poll 方法时,提交上次 poll 返回的所有消息。从顺序上来说,poll 方法的逻辑是先提交上一批消息的位移,再处理下一批消息,因此它能保证不出现消费丢失的情况。
问题描述: 我们系统压测过程中出现下面问题:异常rebalance,而且平均间隔3到5分钟就会触发rebalance,分析日志发现比较严重。错误日志如下:
08-09 11:01:11 131 pool-7-thread-3 ERROR [] -
commit failed
org.apache.kafka.clients.consumer.CommitFailedException: Commit cannot be completed since the group has already rebalanced and assigned the partitions to another member. This means that the time between subsequent calls to poll() was longer than the configured max.poll.interval.ms, which typically implies that the poll loop is spending too much time message processing. You can address this either by increasing the session timeout or by reducing the maximum size of batches returned in poll() with max.poll.records.
at org.apache.kafka.clients.consumer.internals.ConsumerCoordinator.sendOffsetCommitRequest(ConsumerCoordinator.java:713) ~[MsgAgent-jar-with-dependencies.jar:na]
at org.apache.kafka.clients.consumer.internals.ConsumerCoordinator.commitOffsetsSync(ConsumerCoordinator.java:596) ~[MsgAgent-jar-with-dependencies.jar:na]
at org.apache.kafka.clients.consumer.KafkaConsumer.commitSync(KafkaConsumer.java:1218) ~[MsgAgent-jar-with-dependencies.jar:na]
at com.today.eventbus.common.MsgConsumer.run(MsgConsumer.java:121) ~[MsgAgent-jar-with-dependencies.jar:na]
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) [na:1.8.0_161]
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) [na:1.8.0_161]
at java.lang.Thread.run(Thread.java:748) [na:1.8.0_161]
这个错误的意思是,消费者在处理完一批poll的消息后,在同步提交偏移量给broker时报的错。初步分析日志是由于当前消费者线程消费的分区已经被broker给回收了,因为kafka认为这个消费者死了,那么为什么呢?
问题分析:
这里就涉及到问题是消费者在创建时会有一个属性max.poll.interval.ms(默认间隔时间为300s),
该属性意思为kafka消费者在每一轮poll()调用之间的最大延迟,消费者在获取更多记录之前可以空闲的时间量的上限。如果此超时时间期满之前poll()没有被再次调用,则消费者被视为失败,并且分组将重新平衡,以便将分区重新分配给别的成员。
处理重复数据
因为offset此时已经不准确,生产环境不能直接去修改offset偏移量。
所以重新指定了一个消费组(group.id=order_consumer_group),然后指定auto-offset-reset=latest这样我就只需要重启我的服务了,而不需要动kafka和zookeeper了!
#consumerspring.kafka.consumer.group-id=order_consumer_groupspring.kafka.consumer.key-deserializer=org.apache.kafka.common.serialization.StringDeserializerspring.kafka.consumer.value-deserializer=org.apache.kafka.common.serialization.StringDeserializerspring.kafka.consumer.enable-auto-commit=falsespring.kafka.consumer.auto-offset-reset=latest
注:如果你想要消费者从头开始消费某个topic的全量数据,可以重新指定一个全新的group.id=new_group,然后指定auto-offset-reset=earliest即可。