[ 
https://issues.apache.org/jira/browse/HIVE-12266?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14983653#comment-14983653
 ] 

Eugene Koifman commented on HIVE-12266:
---------------------------------------

[~wzheng] I'm not sure it's a good idea to initialize HiveTxnManager in the 
ShutdownHook from Driver().
It may not even be set there yet.  It is first set in Driver.compile() with 
"SessionState.get().initTxnMgr(conf);"
So I think the manager should be passed to the hook there (or even from 
SessionState.initTxnMgr())



> When client exists abnormally, it doesn't release ACID locks
> ------------------------------------------------------------
>
>                 Key: HIVE-12266
>                 URL: https://issues.apache.org/jira/browse/HIVE-12266
>             Project: Hive
>          Issue Type: Bug
>          Components: Transactions
>    Affects Versions: 1.0.0
>            Reporter: Eugene Koifman
>            Assignee: Wei Zheng
>         Attachments: HIVE-12266.1.patch, HIVE-12266.2.patch
>
>
> if you start Hive CLI (locking enabled) and run some command that acquires 
> locks and ^C the shell before command completes the locks for the command 
> remain until they timeout.
> I believe Beeline has the same issue.
> Need to add proper hooks to release locks when command dies. (As much as 
> possible)



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to