[ https://issues.apache.org/jira/browse/HIVE-23110?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17072531#comment-17072531
]
Zoltan Haindrich commented on HIVE-23110:
-----------------------------------------
I think this issue is not connect to the NPE from ReExecDriver; because if that would be there
the catch would have done a [return here|https://github.com/apache/hive/blob/d2ad5b061706a1d3cd55e59c769ed4f2af01cdbe/service/src/java/org/apache/hive/service/cli/operation/SQLOperation.java#L240]
I don't think we have guarantees that after cancel an exception like that may not happen:
* cleanup is synch method (not that it matters)
* it sets state as first step
* and then it start calling driver with .close and .destroy
* as a result of that internally the driver changes to some aborted state
* however..if the job is already near completion it an exception may not happen; and when
the actual thread (Thread-74 in the logs) starts getting out - it doesn't throw an exception;
and that is not in line with the SQLOperation's expectations; which causes the illegal transition
[SQLOperation.cleanup|https://github.com/apache/hive/blob/d2ad5b061706a1d3cd55e59c769ed4f2af01cdbe/service/src/java/org/apache/hive/service/cli/operation/SQLOperation.java#L396]
> Prevent NPE in ReExecDriver if the processing is aborted
> --------------------------------------------------------
>
> Key: HIVE-23110
> URL: https://issues.apache.org/jira/browse/HIVE-23110
> Project: Hive
> Issue Type: Bug
> Components: HiveServer2
> Reporter: Miklos Gergely
> Assignee: Zoltan Haindrich
> Priority: Major
> Attachments: HIVE-23110.01.patch
>
>
> In case of abort the context would be null, and thus the planMapper can not be obtained
from it.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
|