[
https://issues.apache.org/jira/browse/FLINK-12038?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16888946#comment-16888946
]
Andrey Zagrebin commented on FLINK-12038:
-----------------------------------------
True, I see it also in JM logs.
I looped again the test on Travis with the previously suggested fix where the
test waits for yarn app FINISHED state with a timeout.
[https://travis-ci.org/azagrebin/flink/builds/560969859
](the CI fails due to overall timeout because loop has too many iterations but
the test does not fail as previously after couple of iterations)
The waiting does not take long. Then there is no need to kill the app in this
case, only if normal shutdown does not reach FINISHED within a timeout which
would again signal that there is some problem. I think it is a cleaner
approach. I would see the yarn mini cluster shutdown in the @AfterClass test
method as an emergency cleanup.
> YARNITCase stalls on travis
> ---------------------------
>
> Key: FLINK-12038
> URL: https://issues.apache.org/jira/browse/FLINK-12038
> Project: Flink
> Issue Type: Bug
> Components: Deployment / YARN, Tests
> Affects Versions: 1.9.0
> Reporter: Chesnay Schepler
> Assignee: shuai.xu
> Priority: Critical
> Labels: pull-request-available, test-stability
> Fix For: 1.9.0
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> https://travis-ci.org/apache/flink/jobs/511932978
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
