[
https://issues.apache.org/jira/browse/FLINK-9552?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16669883#comment-16669883
]
ASF GitHub Bot commented on FLINK-9552:
---------------------------------------
NicoK opened a new pull request #6971: [FLINK-9552][iterations] fix not syncing
on checkpoint lock before emitting records
URL: https://github.com/apache/flink/pull/6971
## What is the purpose of the change
We need to make sure that concurrent access to the `RecordWriter` is
protected by a lock or otherwise if multiple threads access a `RecordWriter`
instance at the same time, we may get into any type of data corruption which
may not be recognised immediately.
It seems that everything but the `StreamIterationHead` was synchronizing on
the checkpoint lock and hence we should sync here as well.
## Brief change log
- adapt `StreamIterationHead` to synchronize on the checkpoint lock when
emitting records, watermarks, etc.
## Verifying this change
This change is already covered by existing tests, and was tested manually by
adding a check that `RecordWriter` interactions hold the checkpoint lock (by
default, the `RecordWriter` does not know about this lock).
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): **no**
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: **no**
- The serializers: **no**
- The runtime per-record code paths (performance sensitive): **yes**
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Yarn/Mesos, ZooKeeper: **no**
- The S3 file system connector: **no**
## Documentation
- Does this pull request introduce a new feature? **no**
- If yes, how is the feature documented? **not applicable**
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
> NPE in SpanningRecordSerializer during checkpoint with iterations
> -----------------------------------------------------------------
>
> Key: FLINK-9552
> URL: https://issues.apache.org/jira/browse/FLINK-9552
> Project: Flink
> Issue Type: Bug
> Components: Type Serialization System
> Affects Versions: 1.5.0
> Reporter: Truong Duc Kien
> Assignee: Nico Kruber
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.7.0
>
>
> We're encountering NPE intermittently inside SpanningRecordSerializer during
> checkpoint.
>
> {code:java}
> 2018-06-08 08:31:35,741 [ka.actor.default-dispatcher-83] INFO
> o.a.f.r.e.ExecutionGraph IterationSource-22 (44/120)
> (c1b94ef849db0e5fb9fb7b85c17073ce) switched from RUNNING to FAILED.
> java.lang.RuntimeException
> at
> org.apache.flink.streaming.runtime.io.RecordWriterOutput.pushToRecordWriter(RecordWriterOutput.java:110)
> at
> org.apache.flink.streaming.runtime.io.RecordWriterOutput.collect(RecordWriterOutput.java:89)
> at
> org.apache.flink.streaming.runtime.tasks.StreamIterationHead.run(StreamIterationHead.java:91)
> at
> org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:306)
> at org.apache.flink.runtime.taskmanager.Task.run(Task.java:703)
> at java.lang.Thread.run(Thread.java:748)
> Caused by: java.lang.NullPointerException
> at
> org.apache.flink.runtime.io.network.api.serialization.SpanningRecordSerializer.addRecord(SpanningRecordSerializer.java:98)
> at
> org.apache.flink.runtime.io.network.api.writer.RecordWriter.sendToTarget(RecordWriter.java:129)
> at
> org.apache.flink.runtime.io.network.api.writer.RecordWriter.emit(RecordWriter.java:105)
> at
> org.apache.flink.streaming.runtime.io.StreamRecordWriter.emit(StreamRecordWriter.java:81)
> at
> org.apache.flink.streaming.runtime.io.RecordWriterOutput.pushToRecordWriter(RecordWriterOutput.java:107)
> ... 5 more
> 2018-06-08 08:31:35,746 [ka.actor.default-dispatcher-83] INFO
> o.a.f.r.e.ExecutionGraph Job xxx (8a4eaf02c46dc21c7d6f3f70657dbb17) switched
> from state RUNNING to FAILING.
> java.lang.RuntimeException
> at
> org.apache.flink.streaming.runtime.io.RecordWriterOutput.pushToRecordWriter(RecordWriterOutput.java:110)
> at
> org.apache.flink.streaming.runtime.io.RecordWriterOutput.collect(RecordWriterOutput.java:89)
> at
> org.apache.flink.streaming.runtime.tasks.StreamIterationHead.run(StreamIterationHead.java:91)
> at
> org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:306)
> at org.apache.flink.runtime.taskmanager.Task.run(Task.java:703)
> at java.lang.Thread.run(Thread.java:748)
> Caused by: java.lang.NullPointerException
> at
> org.apache.flink.runtime.io.network.api.serialization.SpanningRecordSerializer.addRecord(SpanningRecordSerializer.java:98)
> at
> org.apache.flink.runtime.io.network.api.writer.RecordWriter.sendToTarget(RecordWriter.java:129)
> at
> org.apache.flink.runtime.io.network.api.writer.RecordWriter.emit(RecordWriter.java:105)
> at
> org.apache.flink.streaming.runtime.io.StreamRecordWriter.emit(StreamRecordWriter.java:81)
> at
> org.apache.flink.streaming.runtime.io.RecordWriterOutput.pushToRecordWriter(RecordWriterOutput.java:107)
> ... 5 more
> {code}
> This issue is probably concurrency related, because the revelant Flink code
> seems to have proper null checking
> https://github.com/apache/flink/blob/fa024726bb801fc71cec5cc303cac1d4a03f555e/flink-runtime/src/main/java/org/apache/flink/runtime/io/network/api/serialization/SpanningRecordSerializer.java#L98
> {code:java}
> // Copy from intermediate buffers to current target memory segment
> if (targetBuffer != null) {
> targetBuffer.append(lengthBuffer);
> targetBuffer.append(dataBuffer);
> targetBuffer.commit();
> }
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
