[
https://issues.apache.org/jira/browse/HIVE-15731?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15838833#comment-15838833
]
Siddharth Seth edited comment on HIVE-15731 at 1/25/17 11:41 PM:
-----------------------------------------------------------------
[~sershe] - could you please take a look at the patch.
The main change is to check, ignore and reset the interrupt status in
TezSessionPoolManager.returnSession
The original interrupt hits Hadoop IPC - which catches Interrupt, sets the
interrupt status and throws an IOException. This is during the getDAGStatus
invocation.
[~thejas] - this happens when a ^C is issued in the middle of query execution.
I'm not sure what eventually clears the interrupt status. Maybe the threadpool?
was (Author: sseth):
[~sershe] - could you please take a look at the patch.
The main change is to check, ignore and reset the interrupt status in
TezSessionPoolManager.returnSession
The original interrupt hits Hadoop IPC - which catches Interrupt, sets the
interrupt status and throws an IOException.
[~thejas] - this happens when a ^C is issued in the middle of query execution.
I'm not sure what eventually clears the interrupt status. Maybe the threadpool?
> sessions are not returned to the sessionPool in case of an interrupt
> --------------------------------------------------------------------
>
> Key: HIVE-15731
> URL: https://issues.apache.org/jira/browse/HIVE-15731
> Project: Hive
> Issue Type: Bug
> Reporter: Siddharth Seth
> Assignee: Siddharth Seth
> Priority: Critical
> Attachments: HIVE-15731.01.patch
>
>
> While returning a session to the pool, the interrupt status on the thread
> seems to be set, which causes the pool return to fail.
> The session slot is useless at this point. A HS2 instance configured for a
> single session will stop running queries.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
