[
https://issues.apache.org/jira/browse/TEZ-708?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13932569#comment-13932569
]
Jonathan Eagles commented on TEZ-708:
-------------------------------------
[~bikassaha] and [~sseth] and [~airbots]. I have created a patch based on the
design ideas presented above. At this point, I would appreciate some early
feedback to make sure I am on the correct path before I flesh this patch all
the way out. Here is the essence of the design. Create a TaskScheduler
interface that allows operation with both YarnTaskScheduler
(RemoteTaskScheduler?) and LocalTaskScheduler. Create a bare bones
LocalTaskScheduler that severs the connection to the ResourceManager and strips
out many of the non-essential pieces (delayed containers, reuse). Instead its
purpose is to accept task allocation requests and return fake containers to the
DAGAppMaster via the delegate. One necessary complication added was the
producer-consumer task allocation request queue that would avoid the delegate
synchronization deadlock of hot-wiring the request and the response. Instead
the response is sent in a separate thread. Hope this description help with
understanding the code. Sorry for the class renames, it really muddies up this
small logical change with a rather large patch.
jeagles
> Avoid asking for new containers in uber-local mode
> --------------------------------------------------
>
> Key: TEZ-708
> URL: https://issues.apache.org/jira/browse/TEZ-708
> Project: Apache Tez
> Issue Type: Sub-task
> Reporter: Chen He
> Assignee: Jonathan Eagles
> Attachments: TEZ-708.patch
>
>
> Once we finish this task, the TaskSchedulerEventHandler should not ask for
> new container in the Uber-mode.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
