You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mesos.apache.org by "Dominic Hamon (JIRA)" <ji...@apache.org> on 2014/08/01 00:35:38 UTC

[jira] [Commented] (MESOS-1578) Improve framework rate limiting by imposing the max number of outstanding messages per framework principal

    [ https://issues.apache.org/jira/browse/MESOS-1578?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14081609#comment-14081609 ] 

Dominic Hamon commented on MESOS-1578:
--------------------------------------

Asking a framework to abort when it hits capacity without first giving it a chance to back-off seems aggressive and unfriendly. This is especially the case as operators can set a rate limit on a framework that doesn't have any notion that it's being rate-limited.

Consider sending some messages (non-fatal) to the framework when incoming messages are rate-limited so they can track it and respond before we abort them.

> Improve framework rate limiting by imposing the max number of outstanding messages per framework principal
> ----------------------------------------------------------------------------------------------------------
>
>                 Key: MESOS-1578
>                 URL: https://issues.apache.org/jira/browse/MESOS-1578
>             Project: Mesos
>          Issue Type: Bug
>            Reporter: Yan Xu
>            Assignee: Yan Xu
>
> * Rate limits config takes a configurable *capacity* for each principal.
> * To ensure that Master maintain the message order of a framework it's important that Master sends an FrameworkErrorMessage back to the scheduler to ask it to abort.



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