[
https://issues.apache.org/jira/browse/KAFKA-3065?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15085694#comment-15085694
]
ASF GitHub Bot commented on KAFKA-3065:
---------------------------------------
GitHub user rajinisivaram opened a pull request:
https://github.com/apache/kafka/pull/735
KAFKA-3065: Remove unused topic partitions from RecordAccumulator
Removes unused topic partitions from RecordAccumulator#batches to prevent
the map growing indefinitely. Replaces CopyOnWriteMap with ConcurrentHashMap to
support deletes without double locking. Producer performance tests show no
significant difference.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/rajinisivaram/kafka KAFKA-3065
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/kafka/pull/735.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 #735
----
commit 321aab4207673de26593dcdc00246fe0ca6b6c59
Author: Rajini Sivaram <[email protected]>
Date: 2016-01-05T20:24:27Z
KAFKA-3065: Remove unused topic partitions from RecordAccumulator
----
> Prevent unbounded growth of RecordAccumulator#batches in Kafka producer
> -----------------------------------------------------------------------
>
> Key: KAFKA-3065
> URL: https://issues.apache.org/jira/browse/KAFKA-3065
> Project: Kafka
> Issue Type: Bug
> Components: producer
> Affects Versions: 0.9.0.0
> Reporter: Rajini Sivaram
> Assignee: Rajini Sivaram
>
> Topic partitions added to RecordAccumulator#batches are never removed from
> the map. In a dynamic environment where topics are created and deleted
> frequently, this can result in unbounded growth of the map with unused
> entries. This is a particularly an issue for the Kafka REST service where a
> producer object is used for the lifetime of the service, and the service
> itself does not control the topics to which clients publish messages.
> This is a follow-on defect from KAFKA-2948.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
