[
https://issues.apache.org/jira/browse/KAFKA-4235?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15534556#comment-15534556
]
ASF GitHub Bot commented on KAFKA-4235:
---------------------------------------
GitHub user becketqin opened a pull request:
https://github.com/apache/kafka/pull/1939
KAFKA-4235: Fix the race condition in Sender.initiateClose() that may
result in missing some callbacks.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/becketqin/kafka KAFKA-4235
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/kafka/pull/1939.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #1939
----
----
> Fix the closing order in Sender.initiateClose().
> ------------------------------------------------
>
> Key: KAFKA-4235
> URL: https://issues.apache.org/jira/browse/KAFKA-4235
> Project: Kafka
> Issue Type: Bug
> Affects Versions: 0.10.0.0
> Reporter: Jiangjie Qin
> Assignee: Jiangjie Qin
> Fix For: 0.10.1.0
>
>
> In the initiate close, we need to call RecordAccumulator.close() before set
> the running to false. Otherwise we may miss a callback if the following
> sequence happens:
> 1. Thread 1 call Sender.forceClose(), Sender.forceClosed is set to true,
> Sender.running is set to false.
> 2. Sender thread saw running == false, exited the while loop
> 3. Sender thread saw the forceClose == true and aborted all the existing
> incomplete batches
> 4. Thread 2 append a message into the accumulator.
> 5. Thread 1 closes the accumulator
> 6. Sender thread exits without firing callback for the message in step (4)
> To fix this, we need close the record accumulator before set the running flag
> to false in the sender thread.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
