[ https://issues.apache.org/jira/browse/HIVE-2729?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Todd Lipcon resolved HIVE-2729. ------------------------------- Resolution: Invalid Sorry, it turns out this wasn't Hive's fault after all. The issue was with a corner case of JNI usage - probably the same corner case that caused HIVE-2163. If the first thread to use libhdfs isn't the same thread that started the JVM from C code, then the thread's context classloader will be null. > Hive sets thread's ContextClassLoader to null > --------------------------------------------- > > Key: HIVE-2729 > URL: https://issues.apache.org/jira/browse/HIVE-2729 > Project: Hive > Issue Type: Bug > Affects Versions: 0.8.1 > Reporter: Carl Steinbach > > From HIVE-2163: > https://issues.apache.org/jira/browse/HIVE-2163?focusedCommentId=13188876&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13188876 > {quote} > We came across this same issue in another scenario - it looks like Hive is > somewhere setting the thread's ContextClassLoader to null. I opened > HADOOP-7982 to work around the issue from the Hadoop side, but I think Hive > must have a bug that causes these problems. > ... > Unfortunately the other situation was integrated into some non-open-source > software, so it's not easy to reproduce. It does make use of the Hive > metadata, so my guess is that calling into some SessionState code (as > described above in this ticket) was the culprit. I did verify that the > ContextClassLoader was set to null which is what caused the problem, I just > couldn't tell how it got that way. > {quote} -- 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