[
https://issues.apache.org/jira/browse/HIVE-11272?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14628972#comment-14628972
]
Siddharth Seth commented on HIVE-11272:
---------------------------------------
[~prasanth_j] - I looked at that. Both policies should make use of the priority
field within RuntimeFragmentInfo. Also the parallelism should come from there.
Ran into a case where the ordering was incorrect because a single task from two
vertices had the same start time, even though there was a dependency between
them.
> LLAP: Execution order within LLAP daemons should consider query-specific
> priority assigned to fragments
> -------------------------------------------------------------------------------------------------------
>
> Key: HIVE-11272
> URL: https://issues.apache.org/jira/browse/HIVE-11272
> Project: Hive
> Issue Type: Sub-task
> Reporter: Siddharth Seth
> Assignee: Siddharth Seth
> Priority: Critical
> Fix For: llap
>
>
> It's currently looking at finishable state, start time and vertex
> parallelism. Vertex parallelism can be replaced by upstream parallelism as
> well.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
