[jira] [Commented] (FLINK-9456) Let ResourceManager notify JobManager about failed/killed TaskManagers

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

    [ https://issues.apache.org/jira/browse/FLINK-9456?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16504268#comment-16504268 ] 

ASF GitHub Bot commented on FLINK-9456:
---------------------------------------

GitHub user sihuazhou opened a pull request:

    https://github.com/apache/flink/pull/6132

    [FLINK-9456][Distributed Coordination]Let ResourceManager notify JobManager about failed/killed TaskManagers.

    ## What is the purpose of the change
    
    *Often, the ResourceManager learns faster about TaskManager failures/killings because it directly communicates with the underlying resource management framework. Instead of only relying on the JobManager's heartbeat to figure out that a TaskManager has died, we should additionally send a signal from the ResourceManager to the JobManager if a TaskManager has died. That way, we can react faster to TaskManager failures and recover our running job/s.*
    
    ## Brief change log
    
      - *Add `JobMasterGateway#taskManagerTerminated()` to notify the task manager terminated and do the disconnection there.*
      - *Let the `ResourceManager` to notify JobMaster when the task manager terminated*
    
    ## Verifying this change
    
    - once this approach is verified in general, I will add tests for it.
    
    ## Does this pull request potentially affect one of the following parts:
    
      - Dependencies (does it add or upgrade a dependency): (no)
      - The public API, i.e., is any changed class annotated with `@Public(Evolving)`: (no)
      - The serializers: (no)
      - The runtime per-record code paths (performance sensitive): (no)
      - Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes)
      - The S3 file system connector: (no)
    
    ## Documentation
    
    No

You can merge this pull request into a Git repository by running:

    $ git pull https://github.com/sihuazhou/flink FLINK-9456

Alternatively you can review and apply these changes as the patch at:

    https://github.com/apache/flink/pull/6132.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

    This closes #6132
    
----
commit 652ac037ef3edc75cea0abd4966c2154d6e5fbc0
Author: sihuazhou <su...@...>
Date:   2018-05-10T06:36:27Z

    Let ResourceManager notify JobManager about failed/killed TaskManagers.

----


> Let ResourceManager notify JobManager about failed/killed TaskManagers
> ----------------------------------------------------------------------
>
>                 Key: FLINK-9456
>                 URL: https://issues.apache.org/jira/browse/FLINK-9456
>             Project: Flink
>          Issue Type: Improvement
>          Components: Distributed Coordination
>    Affects Versions: 1.5.0
>            Reporter: Till Rohrmann
>            Assignee: Sihua Zhou
>            Priority: Major
>             Fix For: 1.6.0, 1.5.1
>
>
> Often, the {{ResourceManager}} learns faster about TaskManager failures/killings because it directly communicates with the underlying resource management framework. Instead of only relying on the {{JobManager}}'s heartbeat to figure out that a {{TaskManager}} has died, we should additionally send a signal from the {{ResourceManager}} to the {{JobManager}} if a {{TaskManager}} has died. That way, we can react faster to {{TaskManager}} failures and recover our running job/s.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)