hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chris Douglas (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-4927) Part files on the output filesystem are created irrespective of whether the corresponding task has anything to write there
Date Tue, 03 Feb 2009 00:39:59 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-4927?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12669812#action_12669812
] 

Chris Douglas commented on HADOOP-4927:
---------------------------------------

bq. this feature is a sort of generic across the different output formats and having the framework
support this would be useful.

True. Still, while it is generic functionality, it's neither difficult nor inefficient in
user-space. Absent either of the latter criteria, putting it into the framework seems premature,
at least. If this should be abstracted, wouldn't it make sense as an OutputFormat in lib?
That seems no less brittle than using a framework configuration variable and I've difficulty
seeing this setting scoped to the whole cluster...

bq. I am okay with the current patch in terms of the behavior and the framework support it
adds for lazy creation of the destination output "something" (where "something" is easy to
explain when interpreted as a file).

Unless there's a non-FileOutputFormat use case that's also easy to explain, it remains unclear
that this is the correct abstraction. Creating a setting on FileOutputFormat seems like a
good idea. Whether that's implemented in FileOutputFormat only or via an OuputFormat wrapper
class depends on how generally applicable the abstraction is. Since its motivation is expensive
clutter in HDFS, it's not obvious to me that the latter is justified, let alone tight integration
with the framework.

> Part files on the output filesystem are created irrespective of whether the corresponding
task has anything to write there
> --------------------------------------------------------------------------------------------------------------------------
>
>                 Key: HADOOP-4927
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4927
>             Project: Hadoop Core
>          Issue Type: New Feature
>          Components: mapred
>            Reporter: Devaraj Das
>            Assignee: Jothi Padmanabhan
>             Fix For: 0.21.0
>
>         Attachments: hadoop-4927-v1.patch, hadoop-4927-v2.patch, hadoop-4927.patch
>
>
> When OutputFormat.getRecordWriter is invoked, a part file is created on the output filesystem.
But the created RecordWriter is not used until the OutputCollector.collect call is made by
the task (user's code). This results in empty part files even if the OutputCollector.collect
is never invoked by the corresponding tasks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message