[jira] [Commented] (FLINK-1030) Cleanly separate the Instance Management and Resource Assignment

["ASF GitHub Bot (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-1030?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14117505#comment-14117505 ] 

ASF GitHub Bot commented on FLINK-1030:
---------------------------------------

Github user StephanEwen commented on the pull request:

    https://github.com/apache/incubator-flink/pull/36#issuecomment-54074682
  
    Sorry for the long delay. This looks like cool work, especially unifying the Accumulators and Aggregators :-)
    
    This is interacting quite a bit with the RPC, ExecutionGraph, and JobManager/TaskManager communication (see https://issues.apache.org/jira/browse/FLINK-1030). I am currently reworking that and would like to merge it with that part once it is done.



> Cleanly separate the Instance Management and Resource Assignment
> ----------------------------------------------------------------
>
>                 Key: FLINK-1030
>                 URL: https://issues.apache.org/jira/browse/FLINK-1030
>             Project: Flink
>          Issue Type: Sub-task
>          Components: JobManager
>    Affects Versions: 0.6-incubating
>            Reporter: Stephan Ewen
>            Assignee: Stephan Ewen
>             Fix For: 0.7-incubating
>
>
> Currently, resource assignment is spread between the InstanceManager and the Scheduler.
> I propose to make the InstanceManager only responsible for bookkeeping of live and dead workers.
> The Scheduler and Instances themselves should be responsible for keeping tabs on allocated resources.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)