[
https://issues.apache.org/jira/browse/STORM-1017?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14720207#comment-14720207
]
Sriharsha Chintalapani commented on STORM-1017:
-----------------------------------------------
I think this is harder to fix with IgnoreZkoffsets. Instead we should remove
this option and document if users wants to ignore the zookeeper stored offsets
they can use a new zk node in SpoutConfig or delete the corresponding zknode
where previous offsets are stored.
cc [~ptgoetz] [~parth.brahmbhatt] [~revans2] . Let me know if you see any
better option.
> If ignoreZkOffsets set true,KafkaSpout will reset zk offset when recover from
> failure.
> --------------------------------------------------------------------------------------
>
> Key: STORM-1017
> URL: https://issues.apache.org/jira/browse/STORM-1017
> Project: Apache Storm
> Issue Type: Bug
> Components: storm-kafka
> Reporter: Renkai Ge
>
> when ignoreZkOffsets set true and startOffsetTime =
> kafka.api.OffsetRequest.EarliestTime().
> workers running -> topology shutdown by user and restart -> workers will read
> from earliest time again
> workers running -> one of workers shutdown by accident and supervisor restart
> the worker -> what offset will the restarted worker read from?
> More details on
> https://github.com/apache/storm/pull/493#issuecomment-135783234
> It will cause a lot of unwanted duplicated messages in some scenes.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
