You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mesos.apache.org by "Vinod Kone (JIRA)" <ji...@apache.org> on 2013/05/13 20:43:16 UTC

[jira] [Updated] (MESOS-469) Scheduler driver should call disconnected on master failover

     [ https://issues.apache.org/jira/browse/MESOS-469?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Vinod Kone updated MESOS-469:
-----------------------------

    Description: 
Currently the driver only calls disconnected() when it gets NoMasterDetected() message.

But during a master failover the driver typically only gets NewMasterDetected() and sets an internal variable 'connected = false'. Subsequent driver calls (launchTasks, killTask) are rejected by the driver until it re-registers with a master.

At Twitter, we have seen this window to be as large as ~5s. It would be nice to inform the scheduler that the driver is 'disconnected()' in this situation.

  was:
Currently the driver only calls disconnected() when it gets NoMasterDetected() message.

This means, during a master failover the driver only gets NewMasterDetected() and hence the frameworks are in dark about this disconnection until they re-register with the master.

At Twitter, we have seen this window to be as large as ~5s. So it would be nice to give frameworks a heads up (disconnected()).

    
> Scheduler driver should call disconnected on master failover
> ------------------------------------------------------------
>
>                 Key: MESOS-469
>                 URL: https://issues.apache.org/jira/browse/MESOS-469
>             Project: Mesos
>          Issue Type: Bug
>            Reporter: Vinod Kone
>            Assignee: Vinod Kone
>
> Currently the driver only calls disconnected() when it gets NoMasterDetected() message.
> But during a master failover the driver typically only gets NewMasterDetected() and sets an internal variable 'connected = false'. Subsequent driver calls (launchTasks, killTask) are rejected by the driver until it re-registers with a master.
> At Twitter, we have seen this window to be as large as ~5s. It would be nice to inform the scheduler that the driver is 'disconnected()' in this situation.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira