[jira] [Commented] (HIVE-14042) Potential improvements to abort task handling - tez/llap

["Sergio Peña (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-14042?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15681520#comment-15681520 ] 

Sergio Peña commented on HIVE-14042:
------------------------------------

I'm moving this JIRA out of 2.1.1 release as it is not a blocker nor critical for a 2.1.1 RC version. Feel free to commit it to branch 2.1 if the patch is ready before the release.

> Potential improvements to abort task handling - tez/llap
> --------------------------------------------------------
>
>                 Key: HIVE-14042
>                 URL: https://issues.apache.org/jira/browse/HIVE-14042
>             Project: Hive
>          Issue Type: Improvement
>            Reporter: Siddharth Seth
>            Assignee: Siddharth Seth
>
> Follow up from HIVE-14003.
> - TezProcessor.run() - in case of an abort, throw an interrupt exception instead of falling off. (Falling off cleanly indicates success)
> - Potential cleanup when initialize exits early
> - MapRecordProcessor: MultiMRInput - waitForAllInputsReady can block if interrupts are swallowed elsewhere. Use changes from TEZ-3302.
> - MapRecordProcessor: mergeMapOp handling - this will never be informed about an abort request
> - MapRecordProcessor: Likewise, handling of dummyOp
> - MapRecordProcessor: Abort handling in MergeFileRecordProcessor
> - ReduceRecordProcessor: waitForAllInputsReady. Loop and timed wait after TEZ-3302
> - ReduceRecordProcessor: dummyOp handling



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