kafka 新版本 2.1.0 broker 无故挂起 [英] kafka new version 2.1.0 broker hangs for no reason

问题描述

起初集群中的所有broker都可以正常启动和工作，但有时其中一个broker会遇到问题.并且会出现一些现象:

At first all the brokers in cluster can start and work just fine, but sometimes one of the broker will meet problem. And there are some phenomenon will show up:

• 整个集群挂了，生产者和消费者也不工作，因此网络流量从监视器降为零；
• 使用kafka-topic.sh描述主题消息，每个副本都可以，即使是异常brokerid，zk中的信息也正常；
• 从/proc/sys/fs/file-nr读取的异常broker上的文件描述号逐渐增加
• netstat 代理监听端口 9092 显示大量CLOSE_WAIT"状态

以下是其他broker显示的错误日志，异常broker日志看不到任何异常.

Following is the error log shows from others brokers, while the abnormal broker log can not see any exception.

[2019-04-15 18:10:25,243] INFO [ReplicaFetcher replicaId=0, leaderId=1, fetcherId=0] Error sending fetch request (sessionId=289971597, epoch=1254343) t
o node 1: java.io.IOException: Connection to 1 was disconnected before the response was read. (org.apache.kafka.clients.FetchSessionHandler)
[2019-04-15 18:10:25,244] WARN [ReplicaFetcher replicaId=0, leaderId=1, fetcherId=0] Error in response for fetch request (type=FetchRequest, replicaId=
0, maxWait=500, minBytes=1, maxBytes=10485760, fetchData={atm_error_intf-7=(offset=538244, logStartOffset=5319, maxBytes=1048576, currentLeaderEpoch=Op
tional[13])}, isolationLevel=READ_UNCOMMITTED, toForget=, metadata=(sessionId=289971597, epoch=1254343)) (kafka.server.ReplicaFetcherThread)
java.io.IOException: Connection to 1 was disconnected before the response was read
        at org.apache.kafka.clients.NetworkClientUtils.sendAndReceive(NetworkClientUtils.java:97)
        at kafka.server.ReplicaFetcherBlockingSend.sendRequest(ReplicaFetcherBlockingSend.scala:97)
        at kafka.server.ReplicaFetcherThread.fetchFromLeader(ReplicaFetcherThread.scala:190)
        at kafka.server.AbstractFetcherThread.processFetchRequest(AbstractFetcherThread.scala:241)
        at kafka.server.AbstractFetcherThread.$anonfun$maybeFetch$3(AbstractFetcherThread.scala:130)
        at kafka.server.AbstractFetcherThread.$anonfun$maybeFetch$3$adapted(AbstractFetcherThread.scala:129)
        at scala.Option.foreach(Option.scala:257)
        at kafka.server.AbstractFetcherThread.maybeFetch(AbstractFetcherThread.scala:129)
        at kafka.server.AbstractFetcherThread.doWork(AbstractFetcherThread.scala:111)
        at kafka.utils.ShutdownableThread.run(ShutdownableThread.scala:82)
[2019-04-15 18:10:57,275] WARN [ReplicaFetcher replicaId=0, leaderId=1, fetcherId=0] Error in response for fetch request (type=FetchRequest, replicaId=
0, maxWait=500, minBytes=1, maxBytes=10485760, fetchData={ieg_qsm_guildstatechangereport4pandora-19=(offset=52347859, logStartOffset=38458463, maxBytes=1048576, currentLeaderEpoch=Optional[50]), ieg_qsm_playerreporthighfrequency4pandora-10=(offset=97212897, logStartOffset=65418413, maxBytes=1048576, currentLeaderEpoch=Optional[46]), ieg_qsmtest_guildstatechangereport4pandora-13=(offset=25771, logStartOffset=20917, maxBytes=1048576, currentLeaderEpoch=Optional[46]), __consumer_offsets-10=(offset=0, logStartOffset=0, maxBytes=1048576, currentLeaderEpoch=Optional[57]), ieg_qsmtest_playerreporthighfrequency4pandora-7=(offset=141317, logStartOffset=118323, maxBytes=1048576, currentLeaderEpoch=Optional[45]), __consumer_offsets-25=(offset=0, logStartOffset=0, maxBytes=1048576, currentLeaderEpoch=Optional[44]), ieg_qsmtest_playerlogin-15=(offset=59440, logStartOffset=52149, maxBytes=1048576, currentLeaderEpoch=Optional[55]), dm_pdl_wefeng_findfriend_topic-12=(offset=0, logStartOffset=0, maxBytes=1048576, currentLeaderEpoch=Optional[42]), dm_pdl_wefeng_findfriend_topic_test-0=(offset=0, logStartOffset=0, maxBytes=1048576, currentLeaderEpoch=Optional[47]), ieg_qsmtest_guildstatechangereport4pandora-18=(offset=21042, logStartOffset=16441, maxBytes=1048576, currentLeaderEpoch=Optional[56]), ieg_qsm_playerlogin-1=(offset=27414596, logStartOffset=17328842, maxBytes=1048576, currentLeaderEpoch=Optional[45]), atm_error_intf-7=(offset=538244, logStartOffset=5319, maxBytes=1048576, currentLeaderEpoch=Optional[13]), __consumer_offsets-30=(offset=0, logStartOffset=0, maxBytes=1048576, currentLeaderEpoch=Optional[51]), ieg_qsm_playerreporthighfrequency4pandora-15=(offset=87995984, logStartOffset=53470647, maxBytes=1048576, currentLeaderEpoch=Optional[55]), __consumer_offsets-45=(offset=0, logStartOffset=0, maxBytes=1048576, currentLeaderEpoch=Optional[47]), ieg_qsm_playerlogin-6=(offset=25070198, logStartOffset=16224757, maxBytes=1048576, currentLeaderEpoch=Optional[54]), ieg_qsmtest_playerreporthighfrequency4pandora-12=(offset=141878, logStartOffset=122257, maxBytes=1048576, currentLeaderEpoch=Optional[56]), dm_pdl_wefeng_findfriend_topic-17=(offset=0, logStartOffset=0, maxBytes=1048576, currentLeaderEpoch=Optional[56]), ieg_qsm_guildstatechangereport4pandora-14=(offset=45869398, logStartOffset=27847747, maxBytes=1048576, currentLeaderEpoch=Optional[47]), dm_pdl_wefeng_findfriend_topic_test-5=(offset=0, logStartOffset=0, maxBytes=1048576, currentLeaderEpoch=Optional[57]), atm_error_intf-27=(offset=539432, logStartOffset=5392, maxBytes=1048576, currentLeaderEpoch=Optional[13]), ieg_qsmtest_playerlogin-10=(offset=66712, logStartOffset=55774, maxBytes=1048576, currentLeaderEpoch=Optional[48]), __consumer_offsets-5=(offset=0, logStartOffset=0, maxBytes=1048576, currentLeaderEpoch=Optional[47])}, isolationLevel=READ_UNCOMMITTED, toForget=, metadata=(sessionId=289971597, epoch=INITIAL)) (kafka.server.ReplicaFetcherThread)
java.io.IOException: Connection to 1 was disconnected before the response was read
        at org.apache.kafka.clients.NetworkClientUtils.sendAndReceive(NetworkClientUtils.java:97)
        at kafka.server.ReplicaFetcherBlockingSend.sendRequest(ReplicaFetcherBlockingSend.scala:97)
        at kafka.server.ReplicaFetcherThread.fetchFromLeader(ReplicaFetcherThread.scala:190)
        at kafka.server.AbstractFetcherThread.processFetchRequest(AbstractFetcherThread.scala:241)
        at kafka.server.AbstractFetcherThread.$anonfun$maybeFetch$3(AbstractFetcherThread.scala:130)
        at kafka.server.AbstractFetcherThread.$anonfun$maybeFetch$3$adapted(AbstractFetcherThread.scala:129)
        at scala.Option.foreach(Option.scala:257)
        at kafka.server.AbstractFetcherThread.maybeFetch(AbstractFetcherThread.scala:129)
        at kafka.server.AbstractFetcherThread.doWork(AbstractFetcherThread.scala:111)
        at kafka.utils.ShutdownableThread.run(ShutdownableThread.scala:82)

-Xmx8G -Xms8G，这是jvm堆大小.我只能杀死 -9 实例并重新启动以恢复它.

-Xmx8G -Xms8G, this is the jvm heap size. I only can kill -9 the instance and restart to resume it.

• 版本 kafka_2.12-2.1.0
• Java

推荐答案

这是 kafka 2.1.0 上报告的问题.请查看 https://issues.apache.org/jira/browse/KAFKA-7802 了解详情.它被 https://issues.apache.org/jira/browse/KAFKA 标记为重复-7697 已在 kafka 2.2.0 中修复，因此升级 kafka 版本应该会有所帮助.

This was an issue reported on kafka 2.1.0. Please check https://issues.apache.org/jira/browse/KAFKA-7802 for details. It is marked duplicated by https://issues.apache.org/jira/browse/KAFKA-7697 which is fixed in kafka 2.2.0, so upgrading the kafka version should help.

这篇关于kafka 新版本 2.1.0 broker 无故挂起的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！