简体   繁体   English

卡夫卡重新分配分区工作卡在待定状态

[英]kafka reassign-partitions job stuck in pending

I lost two out of three kafka nodes in my cluster leaving the third one as the leader for all partitions and sole ISR.我在我的集群中丢失了三分之二的 kafka 节点,留下第三个作为所有分区和唯一 ISR 的领导者。 The first two nodes are not coming back so I want to reassign the partitions to two different nodes.前两个节点不会回来,所以我想将分区重新分配给两个不同的节点。

I have tried using both the yahoo kafka-manager tool and also the kafka-reassign-partitions.sh tool.我尝试同时使用 yahoo kafka-manager 工具和 kafka-reassign-partitions.sh 工具。 Both start the job, but in the yahoo kafka-manager the reassignment job shows a status of "pending" forever.两者都开始工作,但在 yahoo kafka-manager 中,重新分配工作永远显示“待定”状态。

  • What are the possible causes of this?造成这种情况的可能原因是什么?
  • Are there additional ways to debug this?还有其他方法可以调试吗?
  • Is it possible to "manually" move the partitions to new nodes?是否可以“手动”将分区移动到新节点?

This could happen when you add brokers with different major version to the cluster, when the API of intra-broker communication has changed between versions, so the controller node can't assign them partitions.当您将具有不同主要版本的代理添加到集群时,当代理内部通信的 API 在版本之间发生变化时,控制器节点无法为它们分配分区时,可能会发生这种情况。 If this is the case, you will need to set a special property in the newly added brokers, to degrade its api version so it will be able to communicate with the old brokers.如果是这种情况,您将需要在新添加的代理中设置一个特殊属性,以降低其 api 版本,以便它能够与旧代理进行通信。 If this is the case, please refer to the upgrade guide of the newer version (eg, for 2.4 - https://kafka.apache.org/24/documentation.html#upgrade )如果是这种情况,请参阅更新版本的升级指南(例如,对于 2.4 - https://kafka.apache.org/24/documentation.html#upgrade

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

相关问题 卡夫卡卡在重新分配分区工具和进度 - Kafka stuck on reassign partitions tool and progress kafka 领导者不可用 - kafka-reassign-partitions - kafka leader unavailable - kafka-reassign-partitions 我可以在 Kafka 上自动重新分配分区吗? - Can I auto-reassign partitions on Kafka? Spark Streaming Kafka作业停留在“处理”阶段 - Spark Streaming Kafka job stuck in 'processing' stage Flink kafka - Flink 作业不向不同分区发送消息 - Flink kafka - Flink job not sending messages to different partitions kafka-reassign-partitions --generate for topic __commit_offsets 给了我奇怪的结果:分区副本只在一个代理上 - kafka-reassign-partitions --generate for topic __commit_offsets gives me strange result: partition replica only on one broker anyway 如果某些Kafka节点时间偏移未同步,则Spark流式传输作业会停止 - Spark streaming job stuck if some Kafka nodes time offset is not synchronized 在 Kafka Streams 中重新平衡期间重建状态存储之前,分区处理卡住 - Partitions processing stuck until state store is rebuilt during rebalancing in Kafka Streams 当我们分配给火花作业的执行者多于kafka主题的分区数时会发生什么 - What happens when we allocate more executors to a spark job than numbers of partitions of a kafka topic 卡夫卡主题与分区 - Kafka topic with Partitions
 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM