Mailing-List: contact issues-help@mesos.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@mesos.apache.org
Date: Tue, 6 Jun 2017 22:01:18 +0000 (UTC)
From: "Michael Park (JIRA)" <jira@apache.org>
To: issues@mesos.apache.org
Message-ID: <JIRA.13077817.1496786389000.11328.1496786478185@Atlassian.JIRA>
In-Reply-To: <JIRA.13077817.1496786389000@Atlassian.JIRA>
References: <JIRA.13077817.1496786389000@Atlassian.JIRA> <JIRA.13077817.1496786389721@jira-lw-us.apache.org>
Subject: [jira] [Assigned] (MESOS-7633) Prevent hierarchical roles from
 being allocated resources from non-HIERARCHICAL_ROLE agents.
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Tue, 06 Jun 2017 22:01:23 -0000


     [ https://issues.apache.org/jira/browse/MESOS-7633?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Michael Park reassigned MESOS-7633:
-----------------------------------

    Assignee: Michael Park

> Prevent hierarchical roles from being allocated resources from non-HIERARCHICAL_ROLE agents.
> --------------------------------------------------------------------------------------------
>
>                 Key: MESOS-7633
>                 URL: https://issues.apache.org/jira/browse/MESOS-7633
>             Project: Mesos
>          Issue Type: Task
>            Reporter: Michael Park
>            Assignee: Michael Park
>
> Since non-{{HIERARCHICAL_ROLE}} agents cannot handle volume creation on hierarchical roles.
> There are a couple of approaches we could have taken here.
> (1) Fine-grained: Offer the resources, and reject requests to create volumes on a hierarchical role.
> (2) Coarse-grained: Don't offer the resources, require that the agent be upgraded before used with a {{HIERARCHICAL_ROLE}} framework.
> The current plan is to pursue (2). A couple of reasons for this. First is that this is a generic behavior that can be expected for any new capability. For example, a {{MULTI_ROLE}} framework does not get offered resources from a non-{{MULTI_ROLE}} agent.
> Another reason is there is no feedback loop for a framework when operations fail currently. I believe that it would be more difficult to debug a situation in which a framework's request is being silently dropped, as opposed to not getting offered resources from the agent to begin with.


--
This message was sent by Atlassian JIRA
(v6.3.15#6346)