[
https://issues.apache.org/jira/browse/HADOOP-7771?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13230188#comment-13230188
]
Daryn Sharp commented on HADOOP-7771:
-------------------------------------
Yes, I wouldn't expect users to know or care about temp files. I don't recall
which of all the copy variants already had this behavior but I standardized it
across the copy commands. Distcp has similar behavior, although I think it
copies into a temp subdir. Maybe it would make sense to mimic what it does?
It's true that users could read the partial files, but I question if users
would expect a partial file in the vast majority of use cases. Let's say a
copy failed, the user didn't check the exit code possibly because 20x exit
codes were unreliable, and then attempted to run a job. The job might
"succeed" to process the partial data but produce erroneous results. Users
might be baffled that re-running the job produces different results. A dev may
go on a wild goose chase and resolve as "works for me" when they can't
reproduce.
If the client is interrupted by a signal, then it often used to leave a temp
file because the signal wasn't breaking out correctly and affecting the
shutdown hooks. I recently fixed that. I've been meaning to
investigation/reproduce/fix a few user reports that temp files are still
sometimes left around. If you do a hard kill (-9), then of course there's
nothing that can be done unless the NN has a notion of this "this file should
be deleted unless I explicitly close it".
What are your thoughts?
> NPE when running hdfs dfs -copyToLocal, -get etc
> ------------------------------------------------
>
> Key: HADOOP-7771
> URL: https://issues.apache.org/jira/browse/HADOOP-7771
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 0.23.0, 0.24.0
> Reporter: John George
> Assignee: John George
> Priority: Blocker
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7771.patch, HADOOP-7771.patch, HADOOP-7771.patch,
> HADOOP-7771.patch, HADOOP-7771.patch, HADOOP-7771.patch
>
>
> NPE when running hdfs dfs -copyToLocal if the destination directory does not
> exist. The behavior in branch-0.20-security is to create the directory and
> copy/get the contents from source.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
