You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Till Rohrmann (JIRA)" <ji...@apache.org> on 2016/08/22 07:56:20 UTC

[jira] [Updated] (FLINK-4352) Implement new JobClient

     [ https://issues.apache.org/jira/browse/FLINK-4352?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Till Rohrmann updated FLINK-4352:
---------------------------------
    Description: 
The {{JobClient}} has to adapted to the new Flip-6 architecture.

1. for yarn mode, jobClient commit SubmitJob/CancelJob to yarn cluster
        when submitJob, yarn cluster create new AppMaster which contains ResourceManager and JobManager for new job
        when cancelJob,  yarn cluster kill the AppMaster and containers
2. for standalone mode, jobClient commit  SubmitJob/CancelJob to LocalDispatcher
         when submitJob, LocalDispatcher spawn new JobManager for job
         when cancelJob, LocalDispatcher send canceJob message to JobManager

  was:
1. for yarn mode, jobClient commit SubmitJob/CancelJob to yarn cluster
        when submitJob, yarn cluster create new AppMaster which contains ResourceManager and JobManager for new job
        when cancelJob,  yarn cluster kill the AppMaster and containers
2. for standalone mode, jobClient commit  SubmitJob/CancelJob to LocalDispatcher
         when submitJob, LocalDispatcher spawn new JobManager for job
         when cancelJob, LocalDispatcher send canceJob message to JobManager


> Implement new JobClient
> -----------------------
>
>                 Key: FLINK-4352
>                 URL: https://issues.apache.org/jira/browse/FLINK-4352
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Cluster Management
>            Reporter: zhangjing
>
> The {{JobClient}} has to adapted to the new Flip-6 architecture.
> 1. for yarn mode, jobClient commit SubmitJob/CancelJob to yarn cluster
>         when submitJob, yarn cluster create new AppMaster which contains ResourceManager and JobManager for new job
>         when cancelJob,  yarn cluster kill the AppMaster and containers
> 2. for standalone mode, jobClient commit  SubmitJob/CancelJob to LocalDispatcher
>          when submitJob, LocalDispatcher spawn new JobManager for job
>          when cancelJob, LocalDispatcher send canceJob message to JobManager



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