[
https://issues.apache.org/jira/browse/KAFKA-15068?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17738591#comment-17738591
]
Chris Egerton commented on KAFKA-15068:
---------------------------------------
Thanks for filing this [~hudeqi], good catch! I agree with your position that
"replication latency reflects the performance of replication and should not be
affected by this abnormal situation".
Let me know if you have any other questions about this; happy to discuss
further here or, if you feel ready, review a PR.
> Incorrect replication Latency may be calculated when the timestamp of the
> record is of type CREATE_TIME
> -------------------------------------------------------------------------------------------------------
>
> Key: KAFKA-15068
> URL: https://issues.apache.org/jira/browse/KAFKA-15068
> Project: Kafka
> Issue Type: Improvement
> Components: mirrormaker
> Reporter: hudeqi
> Assignee: hudeqi
> Priority: Major
>
> When MM2 is used to replicate topics between Kafka clusters, if the timestamp
> of the record of the source cluster is of type CREATE_TIME, then the value of
> timestamp may not be the actual "creation time (that is, append time)", for
> example, the value of timestamp is less than the current time for a long time
> (this is determined by the producer, and often occurs in the online
> environment), at this time, the calculated replication latency will be too
> large.
> I understand that replication latency reflects the performance of replication
> and should not be affected by this abnormal situation.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
