[
https://issues.apache.org/jira/browse/STORM-2554?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16086272#comment-16086272
]
Kristopher Kane commented on STORM-2554:
----------------------------------------
How would you like to work together on this PR? PRs to your fork or patches
here?
Your patch works well. The duplicates and the activate/deactivate problems are
not related to your patch as they are problems in KafkaSpout and the original
Manual* subscription capabilities. It looks like during KafkaSpout.shutdown()
only the ack'd tuples are then committed back to Kafka but I believe that isn't
accounting for other tuples in flight. That fact might be by design for 'at
least once' but I wonder if there is a way to wait for n seconds (based on
topology kill time) to see if the remaining inflight tuples. I don't recall
the same duplicate behavior for in flight tuples when using storm-kafka but
never looked for it specifically.
> Trident Kafka Spout Refactoring to Include Manual Partition Assignment
> ----------------------------------------------------------------------
>
> Key: STORM-2554
> URL: https://issues.apache.org/jira/browse/STORM-2554
> Project: Apache Storm
> Issue Type: Bug
> Components: storm-kafka-client
> Affects Versions: 2.0.0, 1.x
> Reporter: Hugo Louro
> Assignee: Hugo Louro
> Priority: Critical
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Incorporate changes done in STORM-2541 and do some refactoring to internal
> state partition management to make it cleaner and more properly handle
> partitions reassignment.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)