[
https://issues.apache.org/jira/browse/HIVE-16709?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16016807#comment-16016807
]
Gopal V commented on HIVE-16709:
--------------------------------
bq. on abort, will current thread's interrupt flag be set by tez?
Looks like the operator is catching that and eating up the interrupt flag, so
yes, but the processor closing does not get that as the exception gets wrapped
up by the HiveException instead.
> LLAP: Operator abort should traverse the whole tree & cancel all Futures
> ------------------------------------------------------------------------
>
> Key: HIVE-16709
> URL: https://issues.apache.org/jira/browse/HIVE-16709
> Project: Hive
> Issue Type: Bug
> Components: llap
> Affects Versions: 2.1.1, 3.0.0
> Reporter: Gopal V
> Assignee: Gopal V
> Attachments: HIVE-16709.1.patch
>
>
> The logs for pre-empted tasks say
> {code}
> Caused by: org.apache.hadoop.hive.ql.metadata.HiveException: Async
> Initialization failed. abortRequested=false
> at
> org.apache.hadoop.hive.ql.exec.Operator.completeInitialization(Operator.java:456)
> at
> org.apache.hadoop.hive.ql.exec.Operator.initialize(Operator.java:390)
> {code}
> Also for scenarios where TS -> MJ[1] -> MJ[2], the interrupt does not reach
> the MJ[1] operator because an exception is thrown from the
> MJ[2]:completeInitialization() which gets out of initializeChildren() without
> going over the MJ[1] futures.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
