You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@reef.apache.org by "Andrey (JIRA)" <ji...@apache.org> on 2016/08/05 18:23:20 UTC

[jira] [Commented] (REEF-1448) Support for upstream heartbeats

    [ https://issues.apache.org/jira/browse/REEF-1448?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15409860#comment-15409860 ] 

Andrey commented on REEF-1448:
------------------------------

Hi JangHo,
Sorry for delay.
have you put any thoughts into this?
I am thinking one option to address this is to define some observable interface in driver (IJobLifecycleManager, ICancellationToken,... ?)
driver would monitor this interface for IsCancelled status.
and client would provide specific implementation of the interface when submitting the job.


> Support for upstream heartbeats
> -------------------------------
>
>                 Key: REEF-1448
>                 URL: https://issues.apache.org/jira/browse/REEF-1448
>             Project: REEF
>          Issue Type: New Feature
>            Reporter: Markus Weimer
>            Assignee: JangHo Seo
>              Labels: FT
>
> REEF supports heartbeating regarding its own health towards the YARN and Mesos master and the client that spawned the REEF job. What is missing is the reverse: Sometimes, REEF applications need to implement fate sharing with the program / service that launched them. If that service dies, we want the REEF application to shutdown as well.



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