[
https://issues.apache.org/jira/browse/KAFKA-612?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jun Rao updated KAFKA-612:
--------------------------
Attachment: kafka-612.patch
Attach a patch. It introduces a separate method for shutting down empty fetcher
threads and that method will be called at the end of processing a leaderAndIsr
request. Because of this change, a fetcher thread could have an empty fetch
map. To avoid sending empty fetch requests to the broker, the patch uses a
condition to coordinate the fetch. Also fixed a bug when removing items from a
scala map (can't do the removal while iterating the map since the behaviour is
not deterministic).
> move shutting down of fetcher thread out of critical path
> ---------------------------------------------------------
>
> Key: KAFKA-612
> URL: https://issues.apache.org/jira/browse/KAFKA-612
> Project: Kafka
> Issue Type: Bug
> Components: core
> Affects Versions: 0.8
> Reporter: Jun Rao
> Assignee: Jun Rao
> Attachments: kafka-612.patch
>
>
> Shutting down a fetch thread seems to take more than 200ms since we need to
> interrupt the thread. Currently, we shutdown fetcher threads while processing
> a leaderAndIsr request. This can delay some of the partitions to become a
> leader.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
