[
https://issues.apache.org/jira/browse/HADOOP-4490?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12655883#action_12655883
]
Hemanth Yamijala commented on HADOOP-4490:
------------------------------------------
Thanks, Owen. I will take care of that.
Some discussion is required for handling distributed cache (HADOOP-4493).
Firstly, localized files from distributed cache are not localized per job.
Since anything can be passed through distributed cache, I think it should
support the same level of access control as the rest of the files. That is,
they should be changed to be localized per job and subject to the same access
control mechanisms we are using for the rest of the files - like output
directories etc.
I don't think this is a big impact for users as they can't assume the cache to
contain the files they want on the nodes where the task is running. However,
from the system perspective, probably if a lot of users (say working on the
same project that requires the same data files) want to share this but across
multiple jobs, we would be copying only once per node, saving both space and
time. If we modify this to be localized per job, we could lose that advantage,
no ?
Any thoughts on this trade off ?
> Map and Reduce tasks should run as the user who submitted the job
> -----------------------------------------------------------------
>
> Key: HADOOP-4490
> URL: https://issues.apache.org/jira/browse/HADOOP-4490
> Project: Hadoop Core
> Issue Type: Sub-task
> Components: mapred, security
> Reporter: Arun C Murthy
> Assignee: Hemanth Yamijala
>
> Currently the TaskTracker spawns the map/reduce tasks, resulting in them
> running as the user who started the TaskTracker.
> For security and accounting purposes the tasks should be run as the job-owner.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
