[
https://issues.apache.org/jira/browse/KAFKA-3755?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15313028#comment-15313028
]
Jun Rao commented on KAFKA-3755:
--------------------------------
[~imandhan], we can just add a unit test that calls log.append() with
assignOffsets=false and a message with an offset smaller than the log end
offset. The call should fail.
> tightening the offset check in ReplicaFetcherThread
> ---------------------------------------------------
>
> Key: KAFKA-3755
> URL: https://issues.apache.org/jira/browse/KAFKA-3755
> Project: Kafka
> Issue Type: Improvement
> Reporter: Jun Rao
> Assignee: Ishita Mandhan
>
> Currently, in ReplicaFetcherThread.processPartitionData(), we have the
> following code to make sure that the fetchOffset matches the log end offset.
> if (fetchOffset != replica.logEndOffset.messageOffset)
> throw new RuntimeException("Offset mismatch for partition %s: fetched
> offset = %d, log end offset = %d.".format(topicAndPartition, fetchOffset,
> replica.logEndOffset.messageOffset))
> It would be useful to further assert that the first offset in the messageSet
> to be appended to the log is >= than the log end offset.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
