[
https://issues.apache.org/jira/browse/FLINK-13249?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16884990#comment-16884990
]
Yun Gao commented on FLINK-13249:
---------------------------------
Hi [~srichter], I think in this case the key is the Netty thread and the task
thread is mutable blocked and both block state cannot be released.
For the reasons cause this, I future read the code and found that the codes is
similar to let the Netty thread to execute something like
_future.handleAsync(\{routine 1: PartitionProducerStateResponseHandle::new at
Task.java#1090}, Task#executor).thenAccept(\{routine2 :
retriggerPartitionRequest at SingleInputGate#605})_, I think we might intend to
make both routine 1 and routine 2 to be executed with the Task#executor,
however, if the routing 1 executes really fast and when the Netty thread get to
the thenAccept, it found the previous stage has finished, then the Netty thread
will execute routine 2 directly.
> Distributed Jepsen test fails with blocked TaskExecutor
> -------------------------------------------------------
>
> Key: FLINK-13249
> URL: https://issues.apache.org/jira/browse/FLINK-13249
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Network
> Affects Versions: 1.9.0
> Reporter: Till Rohrmann
> Assignee: Stefan Richter
> Priority: Blocker
> Labels: test-stability
> Fix For: 1.9.0
>
> Attachments: jstack_25661_YarnTaskExecutorRunner
>
>
> The distributed Jepsen test which kills {{JobMasters}} started to fail
> recently. From a first glance, it looks as if the {{TaskExecutor's}} main
> thread is blocked by some operation. Further investigation is required.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
