[
https://issues.apache.org/jira/browse/KAFKA-3559?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15581874#comment-15581874
]
ASF GitHub Bot commented on KAFKA-3559:
---------------------------------------
GitHub user enothereska opened a pull request:
https://github.com/apache/kafka/pull/2032
KAFKA-3559: Recycle old tasks when possible
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/enothereska/kafka
KAFKA-3559-onPartitionAssigned
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/kafka/pull/2032.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 #2032
----
commit 28a50430e4136ba75f0d9b957a67f22e7b1e86a0
Author: Eno Thereska <eno.there...@gmail.com>
Date: 2016-10-17T10:46:45Z
Recycle old tasks when possible
----
> Task creation time taking too long in rebalance callback
> --------------------------------------------------------
>
> Key: KAFKA-3559
> URL: https://issues.apache.org/jira/browse/KAFKA-3559
> Project: Kafka
> Issue Type: Improvement
> Components: streams
> Reporter: Guozhang Wang
> Assignee: Eno Thereska
> Labels: architecture
> Fix For: 0.10.2.0
>
>
> Currently in Kafka Streams, we create stream tasks upon getting newly
> assigned partitions in rebalance callback function {code} onPartitionAssigned
> {code}, which involves initialization of the processor state stores as well
> (including opening the rocksDB, restore the store from changelog, etc, which
> takes time).
> With a large number of state stores, the initialization time itself could
> take tens of seconds, which usually is larger than the consumer session
> timeout. As a result, when the callback is completed, the consumer is already
> treated as failed by the coordinator and rebalance again.
> We need to consider if we can optimize the initialization process, or move it
> out of the callback function, and while initializing the stores one-by-one,
> use poll call to send heartbeats to avoid being kicked out by coordinator.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
