You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Ming Liu (JIRA)" <ji...@apache.org> on 2018/12/03 18:21:00 UTC

[jira] [Commented] (KAFKA-7698) Kafka Broker fail to start: ProducerFencedException thrown from producerstatemanager.scala!checkProducerEpoch

    [ https://issues.apache.org/jira/browse/KAFKA-7698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16707621#comment-16707621 ] 

Ming Liu commented on KAFKA-7698:
---------------------------------

The problem is seems coming from appendEndTxnMarker of ProducerStateManager.scala.

ProducerAppendInfo() has this ValidationType passed in. During initial bootstrapping, ValidationType is set to None (instead of Client or All).

So, for Append(), maybeValidateAppend() is called and the check is skipped during loading phase.

But for appendEndTxnMarker(), it seems we should also skip the check. 

> Kafka Broker fail to start: ProducerFencedException thrown from producerstatemanager.scala!checkProducerEpoch 
> --------------------------------------------------------------------------------------------------------------
>
>                 Key: KAFKA-7698
>                 URL: https://issues.apache.org/jira/browse/KAFKA-7698
>             Project: Kafka
>          Issue Type: Bug
>            Reporter: Ming Liu
>            Priority: Major
>
> During our operation of Kafka, we frequently saw this failure: 
>    There was an error in one of the threads during logs loading: org.apache.kafka.common.errors.ProducerFencedException:
> {code:java}
> [06:57:09,697] INFO [ProducerStateManager partition=interaction_events-127] Loading producer state from snapshot file '/data/disk5/kafka/interaction_events-127/00000000092130764817.snapshot' (kafka.log.ProducerStateManager)
> [06:57:09,698] INFO [Log partition=interaction_events-127, dir=/data/disk5/kafka] Completed load of log with 11 segments, log start offset 91975003024 and log end offset 92130764817 in 12701 ms (kafka.log.Log)
> [06:57:09,701] ERROR There was an error in one of the threads during logs loading: org.apache.kafka.common.errors.ProducerFencedException: Producer's epoch is no longer valid. There is probably another producer with a newer epoch. 63 (request epoch), 66 (server epoch) (kafka.log.LogManager)
> [06:57:09,705] INFO [ProducerStateManager partition=client_interaction_events_authorid_enrichment-20] Writing producer snapshot at offset 92418754384 (kafka.log.ProducerStateManager)
> [06:57:09,707] ERROR [KafkaServer id=2] Fatal error during KafkaServer startup. Prepare to shutdown (kafka.server.KafkaServer)
> org.apache.kafka.common.errors.ProducerFencedException: Producer's epoch is no longer valid. There is probably another producer with a newer epoch. 63 (request epoch), 66 (server epoch)
> {code:java}
>  {code}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)