[
https://issues.apache.org/jira/browse/KAFKA-5017?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jason Gustafson updated KAFKA-5017:
-----------------------------------
Resolution: Won't Fix
Status: Resolved (was: Patch Available)
I had thought that this would be OK, but the idempotent producer's dedup check
depends both on the first and last sequence number in the batch, so changing
the former would cause that check to fail. This could result in an incorrect
OutOfSequence error. We could change the logic to only depend on the last
sequence, but it weakens the check and we would need a bit of ugly work to
compute the first offset to include in the duplicate ProduceResponse.
> Consider making baseOffset the first offset in message format v2
> ----------------------------------------------------------------
>
> Key: KAFKA-5017
> URL: https://issues.apache.org/jira/browse/KAFKA-5017
> Project: Kafka
> Issue Type: Sub-task
> Components: clients, core, producer
> Reporter: Ismael Juma
> Assignee: Jason Gustafson
> Fix For: 0.11.0.0
>
>
> Currently baseOffset starts as the first offset of the batch. If the first
> record is removed by compaction, baseOffset doesn't change and it is no
> longer the same as the first offset of the batch. This is inconsistent with
> every other field in the record batch header and it seems like there is no
> longer a reason for this behaviour.
> We should consider simplifying the behaviour so that baseOffset is simply the
> first offset of the record batch. We need to do this before 0.11 or we
> probably won't be able to do it until the next message format version change.
> cc [~hachikuji]
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
