[
https://issues.apache.org/jira/browse/FLINK-26568?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17510402#comment-17510402
]
Yingjie Cao commented on FLINK-26568:
-------------------------------------
Update: I enabled the debug log and reproduced the problem after several days
of testing, but unfortunately, there is no more useful information (no error
and no exception). I notice a new wired phenomenon, only half of the tcp
connection is alive (client to server), which means the client assume that the
connection is good and send a PartitionRequest message to server, but at server
side, the netty processing pipeline has been cleared (or never exist), server
never process the received data (or never receive it), during this process, no
error happens. I can confirm that the partition request message is sent,
because the corresponding log is printed, but no data is received since then.
As illustrated in the following picture, the green box shows the broken tcp
link, only half exists. (The read box shows a normal tcp connection, tasks use
this connection already finished).
I wonder if anyone familiar with this case? Any suggestions?
!image-2022-03-22-18-19-53-171.png!
> BlockingShuffleITCase.testDeletePartitionFileOfBoundedBlockingShuffle timing
> out on Azure
> -----------------------------------------------------------------------------------------
>
> Key: FLINK-26568
> URL: https://issues.apache.org/jira/browse/FLINK-26568
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Network, Runtime / Task
> Affects Versions: 1.15.0
> Reporter: Matthias Pohl
> Priority: Critical
> Labels: test-stability
> Fix For: 1.15.0
>
> Attachments: image-2022-03-22-18-19-53-171.png
>
>
> [This
> build|https://dev.azure.com/mapohl/flink/_build/results?buildId=845&view=logs&j=0a15d512-44ac-5ba5-97ab-13a5d066c22c&t=9a028d19-6c4b-5a4e-d378-03fca149d0b1&l=12865]
> timed out due the test
> {{BlockingShuffleITCase.testDeletePartitionFileOfBoundedBlockingShuffle}} not
> finishing.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
