[
https://issues.apache.org/jira/browse/KAFKA-16662?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17848436#comment-17848436
]
Jianbin Chen commented on KAFKA-16662:
--------------------------------------
After I deleted all of the __cluster_metadata-0, the problem did not occur when
I started the cluster, but all my topic information was lost. Fortunately, this
is just an offline test environment cluster. According to the phenomenon, it is
certain that the incompatibility between the 3.5 version of metadata and the
3.7 version caused this problem. This makes me dare not try to smoothly upgrade
the cluster. In the past, when using zk, upgrading the broker would never cause
similar problems!
> UnwritableMetadataException: Metadata has been lost
> ---------------------------------------------------
>
> Key: KAFKA-16662
> URL: https://issues.apache.org/jira/browse/KAFKA-16662
> Project: Kafka
> Issue Type: Bug
> Affects Versions: 3.7.0
> Environment: Docker Image (bitnami/kafka:3.7.0)
> via Docker Compose
> Reporter: Tobias Bohn
> Priority: Major
> Attachments: log.txt
>
>
> Hello,
> First of all: I am new to this Jira and apologize if anything is set or
> specified incorrectly. Feel free to advise me.
> We currently have an error in our test system, which unfortunately I can't
> solve, because I couldn't find anything related to it. No solution could be
> found via the mailing list either.
> The error occurs when we want to start up a node. The node runs using Kraft
> and is both a controller and a broker. The following error message appears at
> startup:
> {code:java}
> kafka | [2024-04-16 06:18:13,707] ERROR Encountered fatal fault: Unhandled
> error initializing new publishers
> (org.apache.kafka.server.fault.ProcessTerminatingFaultHandler)
> kafka | org.apache.kafka.image.writer.UnwritableMetadataException: Metadata
> has been lost because the following could not be represented in metadata
> version 3.5-IV2: the directory assignment state of one or more replicas
> kafka | at
> org.apache.kafka.image.writer.ImageWriterOptions.handleLoss(ImageWriterOptions.java:94)
> kafka | at
> org.apache.kafka.metadata.PartitionRegistration.toRecord(PartitionRegistration.java:391)
> kafka | at org.apache.kafka.image.TopicImage.write(TopicImage.java:71)
> kafka | at
> org.apache.kafka.image.TopicsImage.write(TopicsImage.java:84)
> kafka | at
> org.apache.kafka.image.MetadataImage.write(MetadataImage.java:155)
> kafka | at
> org.apache.kafka.image.loader.MetadataLoader.initializeNewPublishers(MetadataLoader.java:295)
> kafka | at
> org.apache.kafka.image.loader.MetadataLoader.lambda$scheduleInitializeNewPublishers$0(MetadataLoader.java:266)
> kafka | at
> org.apache.kafka.queue.KafkaEventQueue$EventContext.run(KafkaEventQueue.java:127)
> kafka | at
> org.apache.kafka.queue.KafkaEventQueue$EventHandler.handleEvents(KafkaEventQueue.java:210)
> kafka | at
> org.apache.kafka.queue.KafkaEventQueue$EventHandler.run(KafkaEventQueue.java:181)
> kafka | at java.base/java.lang.Thread.run(Thread.java:840)
> kafka exited with code 0 {code}
> We use Docker to operate the cluster. The error occurred while we were trying
> to restart a node. All other nodes in the cluster are still running correctly.
> If you need further information, please let us know. The complete log is
> attached to this issue.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
