You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mesos.apache.org by "Yifan Gu (JIRA)" <ji...@apache.org> on 2014/06/04 10:07:01 UTC

[jira] [Commented] (MESOS-886) Slave should wait until resources are isolated before launching tasks

    [ https://issues.apache.org/jira/browse/MESOS-886?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14017476#comment-14017476 ] 

Yifan Gu commented on MESOS-886:
--------------------------------

Hey guys.
I am interested in fixing this issue.
My plan is to catch the future returned by containerizer->update(), and set up a defered call __runTask() which check the future's state before sending the RunTaskMessage.

Any advice and questions are highly welcome! Thanks!

> Slave should wait until resources are isolated before launching tasks
> ---------------------------------------------------------------------
>
>                 Key: MESOS-886
>                 URL: https://issues.apache.org/jira/browse/MESOS-886
>             Project: Mesos
>          Issue Type: Bug
>          Components: isolation, slave
>    Affects Versions: 0.14.0
>            Reporter: Ian Downes
>            Priority: Minor
>
> The slave dispatches to the isolator to update resources and then sends RunTaskMessage to the executor without waiting for the update to complete. 
> This race could, for example, lead to the task using too much RAM (including file cache) and then being OOM killed whenever the resource update completes.



--
This message was sent by Atlassian JIRA
(v6.2#6252)