Harsh J created OOZIE-822:
-----------------------------

             Summary: Document that the Oozie Hive action creates a 
"hive-site.xml" file in the task's working directory
                 Key: OOZIE-822
                 URL: https://issues.apache.org/jira/browse/OOZIE-822
             Project: Oozie
          Issue Type: Improvement
          Components: action
    Affects Versions: 3.2.0
            Reporter: Harsh J
            Assignee: Harsh J
            Priority: Trivial


In HiveMain.java, there is some code that does 'OutputStream os = new 
FileOutputStream("hive-site.xml");'. This indicates that running HiveMain will 
make it create a hive-site.xml in the launcher-mappers' working directory 
itself.

Now if a user has <job-xml> set to a HDFS file named "hive-site.xml" too, it 
will create a symlink of this file in the working directory before the launcher 
task begins. In insecure mode this isn't an issue cause the LauncherMapper runs 
as the same user as the TT and will overwrite this file when it runs the above 
mentioned code. However, since distributed cache symlinks are owned by the 
TT-running user (mapred, say), in secure MR mode the LauncherMapper runs as the 
actual user and the code runs into a permission issue as it can't overwrite a 
file it does not own.

We should hence document that one should not pass such a filename into the 
workflow that would make it symlink to work directory, to help avoid such a 
conflict.

I'll think of the right words to put these in crisply and open up a review 
request.

--
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