[
https://issues.apache.org/jira/browse/FLINK-4905?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15708105#comment-15708105
]
Andrew Efimov edited comment on FLINK-4905 at 11/30/16 10:38 AM:
-----------------------------------------------------------------
Hi [~tzulitai]
Thanks for reply.
It is very difficult to reproduce this state of concurrency model. But we
should make the model more transparent and predictable.
I have created diagram that describes behaviour (used free idea plugin
plantuml):
see attached file Kafka08Fetcher.png
There are two thread which can use the same Kafka08Fetcher and for each we
should make a guarantee that curator client will be working at the moment of
using.
And I would suggest two approaches, please take a look at the draft:
1. If each StreamTask has its Kafka08Fetcher then we can create a flag that
will indicate commit handler is used or not
(or we can create atomicInteger to increment and decrement count of thread
which use handler)
https://github.com/BrainLogic/flink/commit/ba327dc5c991d12366ab16ea6d8e707238b2e79c
2. using checkpoint lock from sourceContext
https://github.com/BrainLogic/flink/commit/abe16442c5dcf294b3c2bd28142765fe28515f8c
Let's discuss and then I'm going to do tests.
was (Author: andrew efimov):
Hi [~tzulitai]
Thanks for reply.
It is very difficult to reproduce this state of concurrency model. But we
should make the model more transparent and predictable.
I have created diagram that describes behaviour (used free idea plugin
plantuml):
see attached file Kafka08Fetcher.png
There are two thread which can use the same Kafka08Fetcher and for each we
should make a guarantee that curator client will be working at the moment of
using.
And I would suggest two approaches, please take a look at the draft:
1. If each StreamTask has its Kafka08Fetcher then we can create a flag that
will indicate commit handler is used or not
(or we can create atomicInteger to increment and decrement count of thread
which use handler)
https://github.com/apache/flink/commit/ba327dc5c991d12366ab16ea6d8e707238b2e79c
2. using checkpoint lock from sourceContext
https://github.com/BrainLogic/flink/commit/abe16442c5dcf294b3c2bd28142765fe28515f8c
Let's discuss and then I'm going to do tests.
> Kafka test instability IllegalStateException: Client is not started
> -------------------------------------------------------------------
>
> Key: FLINK-4905
> URL: https://issues.apache.org/jira/browse/FLINK-4905
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector
> Reporter: Robert Metzger
> Labels: test-stability
> Attachments: Kafka08Fetcher.png
>
>
> The following travis build
> (https://s3.amazonaws.com/archive.travis-ci.org/jobs/170365439/log.txt)
> failed because of this error
> {code}
> 08:17:11,239 INFO org.apache.flink.runtime.jobmanager.JobManager
> - Status of job 33ebdc0e7c91be186d80658ce3d17069 (Read some records to
> commit offsets to Kafka) changed to FAILING.
> java.lang.RuntimeException: Error while confirming checkpoint
> at org.apache.flink.runtime.taskmanager.Task$4.run(Task.java:1040)
> at
> java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:471)
> at java.util.concurrent.FutureTask.run(FutureTask.java:262)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> at java.lang.Thread.run(Thread.java:745)
> Caused by: java.lang.IllegalStateException: Client is not started
> at
> org.apache.flink.shaded.com.google.common.base.Preconditions.checkState(Preconditions.java:173)
> at
> org.apache.curator.CuratorZookeeperClient.getZooKeeper(CuratorZookeeperClient.java:113)
> at
> org.apache.curator.utils.EnsurePath$InitialHelper$1.call(EnsurePath.java:148)
> at org.apache.curator.RetryLoop.callWithRetry(RetryLoop.java:107)
> at
> org.apache.curator.utils.EnsurePath$InitialHelper.ensure(EnsurePath.java:141)
> at org.apache.curator.utils.EnsurePath.ensure(EnsurePath.java:99)
> at
> org.apache.flink.streaming.connectors.kafka.internals.ZookeeperOffsetHandler.setOffsetInZooKeeper(ZookeeperOffsetHandler.java:133)
> at
> org.apache.flink.streaming.connectors.kafka.internals.ZookeeperOffsetHandler.prepareAndCommitOffsets(ZookeeperOffsetHandler.java:93)
> at
> org.apache.flink.streaming.connectors.kafka.internals.Kafka08Fetcher.commitInternalOffsetsToKafka(Kafka08Fetcher.java:341)
> at
> org.apache.flink.streaming.connectors.kafka.FlinkKafkaConsumerBase.notifyCheckpointComplete(FlinkKafkaConsumerBase.java:421)
> at
> org.apache.flink.streaming.api.operators.AbstractUdfStreamOperator.notifyOfCompletedCheckpoint(AbstractUdfStreamOperator.java:229)
> at
> org.apache.flink.streaming.runtime.tasks.StreamTask.notifyCheckpointComplete(StreamTask.java:571)
> at org.apache.flink.runtime.taskmanager.Task$4.run(Task.java:1035)
> ... 5 more
> 08:17:11,241 INFO org.apache.flink.runtime.taskmanager.Task
> - Attempting to cancel task Source: Custom Source -> Map -> Map -> Sink:
> Unnamed (1/3)
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
