[
https://issues.apache.org/jira/browse/TEZ-971?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13944691#comment-13944691
]
Siddharth Seth commented on TEZ-971:
------------------------------------
At the moment, a fatal error results in the process itself being killed - via
the state machines in the AM. Eventually, yes - I think we need interrupt the
thread. Not all Inputs / Outputs respect interrupts at the moment though -
which is why we rely on killing them.
TEZ-919 will make this better - but I think more work will be needed after that
as well.
> Shuffle should report errors early
> ----------------------------------
>
> Key: TEZ-971
> URL: https://issues.apache.org/jira/browse/TEZ-971
> Project: Apache Tez
> Issue Type: Bug
> Reporter: Siddharth Seth
> Assignee: Siddharth Seth
> Priority: Critical
> Attachments: TEZ-971.1.txt
>
>
> Currently, shuffle requires a getNext / isAvailable method to be called
> before it reports exceptions. This needs to be done early.
> Related TEZ-919, but planning on fixing this first.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
