[ 
https://issues.apache.org/jira/browse/HDFS-3083?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Aaron T. Myers updated HDFS-3083:
---------------------------------

    Attachment: HDFS-3083-combined.patch

Here's a patch which addresses the issue. It includes changes in both HDFS and 
Common projects, so test-patch isn't going to work. I can create separate JIRAs 
if folks want, but I figure reviewing it would be easier as a single patch.

No tests are included since security has to be enabled to verify the fix. To 
test it out, I ran the DT test script attached to HDFS-2904, with the following 
extra test case appended:

{code}
# Token issued by nn2 should work when nn2 still active
kinit -k -t ~/keytabs/$ADMIN.keytab $ADMIN/simon
kinit -R
hdfs haadmin -failover nn1 nn2
rm -f /tmp/token
hdfs fetchdt --renewer $RENEWER /tmp/token
kdestroy
HADOOP_TOKEN_FILE_LOCATION=/tmp/token hadoop fs -ls /
{code}

All of the tests in the test script passed with this patch applied. The above 
test fails without the patch, and passes with it. I also successfully ran some 
MR jobs with the second-listed NN in the active state, and confirmed that 
everything worked as expected.
                
> HA+security: failed to run a mapred job from yarn after a manual failover
> -------------------------------------------------------------------------
>
>                 Key: HDFS-3083
>                 URL: https://issues.apache.org/jira/browse/HDFS-3083
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: ha, security
>    Affects Versions: 0.24.0, 0.23.3
>            Reporter: Mingjie Lai
>            Assignee: Aaron T. Myers
>            Priority: Critical
>             Fix For: 0.24.0, 0.23.3
>
>         Attachments: HDFS-3083-combined.patch
>
>
> Steps to reproduce:
> - turned on ha and security
> - run a mapred job, and wait to finish
> - failover to another namenode
> - run the mapred job again, it fails. 

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to