[
https://issues.apache.org/jira/browse/KAFKA-804?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13602457#comment-13602457
]
Jun Rao commented on KAFKA-804:
-------------------------------
Yes, the log code is a bit more complicated now to address a few recent issues
that we found :
kafka-802: flush interval based on compressed message set
kafka-765: corrupted messages cause broker to shut down
kafka-767: message size check needs to be done after offset assignment
My main concern of not addressing this in 0.8 is that this makes it hard to do
index verification in our system tests.
Indexing the first message seems to make the most intuitive sense. Yes, we do
need to patch DumpLogSegments to support deep message iteration, in order to do
index verification.
> Incorrect index in the log of a follower
> ----------------------------------------
>
> Key: KAFKA-804
> URL: https://issues.apache.org/jira/browse/KAFKA-804
> Project: Kafka
> Issue Type: Bug
> Components: core
> Affects Versions: 0.8
> Reporter: Jun Rao
> Assignee: Jun Rao
> Priority: Blocker
> Attachments: kafka-804.patch
>
>
> In the follower, in log.append(), we use the offset of the 1st compressed
> message as the offset for appending index entries. This means that the index
> is pointing to an earlier position in the file than it should. Instead, it
> should use the offset of the 1st uncompressed message for appending index.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
