[
https://issues.apache.org/jira/browse/HADOOP-4493?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12741439#action_12741439
]
Hemanth Yamijala commented on HADOOP-4493:
------------------------------------------
Had a chat with Owen to get more details. Here is the summary:
- To download the files from distributed cache, we will use the user's JobConf
for getting the file system handle. Owen feels any changes to Hadoop security
will be abstracted here and hence we will not be affected.
- In this JIRA, we'll manage a per user cache where the downloaded cache files
will be managed securely as done in MAPREDUCE-842 (ownership set to user, group
ownership to the tasktracker user, 570 permissions).
- The per user cache will be under $ttroot/users/$user/archive
- Owen suggested we can change the jobcache also to be under
$ttroot/users/$user/jobcache as part of this jira. (However, we will not change
the direction of MAPREDUCE-842 which is quite close to the finish line now)
- There'll be symlinks from the task directories to the distributed cache
files.
- The current paging policy which sets a soft limit on the size of the
distributed cache on disk will remain unchanged. We discussed there could be
potentially work in the future to make this a hard limit, which could even have
impacts on scheduling in the sense that the scheduler should schedule tasks to
be within the limits). But this is certainly the scope of another jira.
> Localized files from DistributedCache should have right access-control
> ----------------------------------------------------------------------
>
> Key: HADOOP-4493
> URL: https://issues.apache.org/jira/browse/HADOOP-4493
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: security
> Reporter: Arun C Murthy
>
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
