[
https://issues.apache.org/jira/browse/TEZ-3336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15375691#comment-15375691
]
Siddharth Seth commented on TEZ-3336:
-------------------------------------
[~jlowe] - InputInitializer events and VMEvents short circuit the DAG
definition to some extent. Pretty much any task can send these events to any
Vertex since there's no Control Plane definition to restrict this.
The Hive processor is supposed to target these events at specific Vertices
which know how to handle them. In fact I'm not sure Hive uses
MRInputSplitGeneartor at all anymore. It has it's own SplitGenerator which is
based on MRInputSplitGeneartor - and knows how to handle these events for
partition pruning. This sounds like a Hive bug to me.
> Hive map-side join job sometimes fails with ROOT_INPUT_INIT_FAILURE
> -------------------------------------------------------------------
>
> Key: TEZ-3336
> URL: https://issues.apache.org/jira/browse/TEZ-3336
> Project: Apache Tez
> Issue Type: Bug
> Affects Versions: 0.7.1
> Reporter: Jason Lowe
>
> When Hive does a map-side join it can generate a DAG where a vertex has two
> inputs, one from an upstream task and another using MRInputAMSplitGenerator.
> If it takes a while for MRInputAMSplitGenerator to compute the splits and one
> of the tasks for the other upstream vertex completes then the job can fail
> with an error since MRInputAMSplitGenerator does not expect to receive any
> events.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
