[
https://issues.apache.org/jira/browse/HDFS-9515?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15060951#comment-15060951
]
Arpit Agarwal commented on HDFS-9515:
-------------------------------------
+1 thanks for this improvement [~jojochuang]. When cluster/fs are class members
perhaps it also makes sense to set {{cluster = null}} after invoking
{{cluster.shutdown()}}, so @After invocations of subsequent test methods don't
call shutdown twice on the same object.
> NPE in TestDFSZKFailoverController due to binding exception in
> MiniDFSCluster.initMiniDFSCluster()
> --------------------------------------------------------------------------------------------------
>
> Key: HDFS-9515
> URL: https://issues.apache.org/jira/browse/HDFS-9515
> Project: Hadoop HDFS
> Issue Type: Bug
> Environment: Jenkins
> Reporter: Wei-Chiu Chuang
> Assignee: Wei-Chiu Chuang
> Priority: Minor
> Attachments: HDFS-9515.001.patch, HDFS-9515.002.patch
>
>
> If MiniDFSCluster constructor throws an exception, the cluster object is not
> assigned, so shutdown() call not be called on the object.
> I saw in a recent Jenkins job where binding error threw an exception, and
> later on the NPE of cluster.shutdown() hid the real cause of the test failure.
> HDFS-9333 has a patch that fixes the bind error.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
