[
https://issues.apache.org/jira/browse/HDFS-6312?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13985714#comment-13985714
]
Daryn Sharp commented on HDFS-6312:
-----------------------------------
I found the bug by just looking at the code. {{resetStateToFailOver}} nulls
out the internal token and clears the init flag in {{TokenAspect}}. The next
operation that calls {{getDelegationToken()}} to build the auth params for the
url will attempt to acquire a new token.
HADOOP-10519 is a different issue. The HA support in webhdfs has many flaws.
> WebHdfs HA failover is broken on secure clusters
> ------------------------------------------------
>
> Key: HDFS-6312
> URL: https://issues.apache.org/jira/browse/HDFS-6312
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: webhdfs
> Affects Versions: 3.0.0, 2.4.0
> Reporter: Daryn Sharp
> Priority: Blocker
>
> When webhdfs does a failover, it blanks out the delegation token. This will
> cause subsequent operations against the other NN to acquire a new token.
> Tasks cannot acquire a token (no kerberos credentials) so jobs will fail.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
