[
https://issues.apache.org/jira/browse/KAFKA-2886?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15096726#comment-15096726
]
ASF GitHub Bot commented on KAFKA-2886:
---------------------------------------
GitHub user hachikuji opened a pull request:
https://github.com/apache/kafka/pull/767
KAFKA-2886: handle sink task rebalance failures by stopping worker task
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/hachikuji/kafka KAFKA-2886
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/kafka/pull/767.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 #767
----
commit eddfb3afc46ca9b0747ed57aca8dd5b50e4d4519
Author: Jason Gustafson <ja...@confluent.io>
Date: 2016-01-13T18:09:10Z
KAFKA-2886: handle sink task rebalance failures by stopping worker task
----
> WorkerSinkTask doesn't catch exceptions from rebalance callbacks
> ----------------------------------------------------------------
>
> Key: KAFKA-2886
> URL: https://issues.apache.org/jira/browse/KAFKA-2886
> Project: Kafka
> Issue Type: Bug
> Components: copycat
> Reporter: Ewen Cheslack-Postava
> Assignee: Jason Gustafson
>
> WorkerSinkTask exposes rebalance callbacks to tasks by invoking
> onPartitionsRevoked and onPartitionsAssigned on the task. However, these
> aren't guarded by try/catch blocks, so they can propagate the errors up to
> the consumer:
> {quote}
> [2015-11-24 15:52:24,071] ERROR User provided listener
> org.apache.kafka.connect.runtime.WorkerSinkTask$HandleRebalance failed on
> partition assignment:
> (org.apache.kafka.clients.consumer.internals.ConsumerCoordinator)
> java.lang.UnsupportedOperationException
> at
> java.util.Collections$UnmodifiableCollection.clear(Collections.java:1094)
> at
> io.confluent.connect.hdfs.DataWriter.onPartitionsAssigned(DataWriter.java:207)
> at
> io.confluent.connect.hdfs.HdfsSinkTask.onPartitionsAssigned(HdfsSinkTask.java:103)
> at
> org.apache.kafka.connect.runtime.WorkerSinkTask$HandleRebalance.onPartitionsAssigned(WorkerSinkTask.java:369)
> at
> org.apache.kafka.clients.consumer.internals.ConsumerCoordinator.onJoinComplete(ConsumerCoordinator.java:189)
> at
> org.apache.kafka.clients.consumer.internals.AbstractCoordinator.ensureActiveGroup(AbstractCoordinator.java:227)
> at
> org.apache.kafka.clients.consumer.internals.ConsumerCoordinator.ensurePartitionAssignment(ConsumerCoordinator.java:306)
> at
> org.apache.kafka.clients.consumer.KafkaConsumer.pollOnce(KafkaConsumer.java:861)
> at
> org.apache.kafka.clients.consumer.KafkaConsumer.poll(KafkaConsumer.java:829)
> at
> org.apache.kafka.connect.runtime.WorkerSinkTask.poll(WorkerSinkTask.java:171)
> at
> org.apache.kafka.connect.runtime.WorkerSinkTaskThread.iteration(WorkerSinkTaskThread.java:90)
> at
> org.apache.kafka.connect.runtime.WorkerSinkTaskThread.execute(WorkerSinkTaskThread.java:58)
> at
> org.apache.kafka.connect.util.ShutdownableThread.run(ShutdownableThread.java:82)
> [2015-11-24 15:52:24,477] INFO Cannot acquire lease on WAL
> hdfs://worker4:9000/logs/test/0/log (io.confluent.connect.hdfs.wal.FSWAL)
> {quote}
> This actually currently works ok for onPartitionsAssigned because the
> callback is the last thing invoked. For onPartitionsRevoked, it causes
> offsets to not be committed and the current message batch being processed to
> not be cleared. Additionally, we may need to do something more to clean up,
> e.g. the task may need to stop processing data entirely since the task may
> now be in a bad state.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
