[
https://issues.apache.org/jira/browse/HADOOP-6554?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12833753#action_12833753
]
Steve Loughran commented on HADOOP-6554:
----------------------------------------
# I think the check should maybe include one for null-ness, not just
is-running. Just to be sure
# I can envisage a fairly straightforward test for this problem too
# any call to System.exit() is bad and will only get picked up and rejected by
a security manager. Best to make handing explicit -and have a separate bugrep
for it.
> DelegationTokenSecretManager lifecycle is inconsistent
> ------------------------------------------------------
>
> Key: HADOOP-6554
> URL: https://issues.apache.org/jira/browse/HADOOP-6554
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 0.22.0
> Reporter: Todd Lipcon
> Assignee: Todd Lipcon
> Attachments: hadoop-6554.txt, hdfs-967.txt
>
>
> FSNamesystem.dtSecretManager is initialized in FSNamesystem.initialize(), but
> not started until activate(). In FSN.stop() it calls
> dtSecretManager.stopThreads so long as it's non-null, but doesn't check that
> the threads have started (ie activate() may never have been called). This
> causes an NPE in stopThreads()
> To fix this, stopThreads should check tokenRemoverThread against null before
> interrupting it.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
