crunch-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Micah Whitacre (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CRUNCH-394) Ensure that the Broadcast configuration in SparkRuntimeContext is never null
Date Fri, 16 May 2014 10:38:40 GMT

    [ https://issues.apache.org/jira/browse/CRUNCH-394?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13998894#comment-13998894
] 

Micah Whitacre commented on CRUNCH-394:
---------------------------------------

+1

> Ensure that the Broadcast configuration in SparkRuntimeContext is never null
> ----------------------------------------------------------------------------
>
>                 Key: CRUNCH-394
>                 URL: https://issues.apache.org/jira/browse/CRUNCH-394
>             Project: Crunch
>          Issue Type: Bug
>            Reporter: Josh Wills
>             Fix For: 0.10.0, 0.8.3
>
>         Attachments: CRUNCH-394.patch
>
>
> It looks like it's possible for Spark to recompute some RDDs that get reloaded between
output runs. When this happens, the Broadcast<byte[]> field that contains the serialized
Configuration object is null b/c we haven't declared an output RDD to run against yet. I added
a change to ensure that the serialized Configuration object is always defined, even before
we generate any output, to ensure that we don't throw an NPE.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message