[jira] [Commented] (MESOS-6338) Support --revocable_cpu_low_priority flag for docker containerizer

["Jie Yu (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MESOS-6338?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15556038#comment-15556038 ] 

Jie Yu commented on MESOS-6338:
-------------------------------

Sounds good. Keep in mind that docker containerizer will receive less support than mesos containerizer in the future, and new features (e.g., pod, gpus) will go to mesos containerizer first typically.

> Support --revocable_cpu_low_priority flag for docker containerizer
> ------------------------------------------------------------------
>
>                 Key: MESOS-6338
>                 URL: https://issues.apache.org/jira/browse/MESOS-6338
>             Project: Mesos
>          Issue Type: Improvement
>          Components: containerization
>            Reporter: Kunal Thakar
>
> The mesos containerizer supports setting lower shares for revocable tasks by passing --revocable_cpu_low_priority to the mesos agent. This flag is only supported for mesos containerizer, but I don't see a reason why the behavior can't be replicated for the docker containerizer. 
> On setting the flag, CPU shares assigned to revocable tasks are lower than normal tasks (https://github.com/apache/mesos/blob/master/src/slave/containerizer/mesos/isolators/cgroups/subsystems/cpu.cpp#L83). This does not happen in the docker containerizer (https://github.com/apache/mesos/blob/master/src/slave/containerizer/docker.cpp#L1517), but it can be easily replicated there. 
> I can send a patch if this is acceptable.



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