aurora-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "David Robinson (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (AURORA-662) Improve Rack Limit Default
Date Mon, 08 Sep 2014 18:40:28 GMT

    [ https://issues.apache.org/jira/browse/AURORA-662?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14125891#comment-14125891
] 

David Robinson commented on AURORA-662:
---------------------------------------

I don't see why this should be configurable, the option shouldn't even exist. The scheduler
should be aware of all failure domains (server, rack, power distribution, DC) and should schedule
jobs to be resilient to these failures.

> Improve Rack Limit Default
> --------------------------
>
>                 Key: AURORA-662
>                 URL: https://issues.apache.org/jira/browse/AURORA-662
>             Project: Aurora
>          Issue Type: Story
>          Components: Scheduler
>            Reporter: Zameer Manji
>            Assignee: Zameer Manji
>
> If a rack limit for a job is unspecified, the scheduler places a default of 1. This default
is naive and can make large jobs impossible to schedule.
> We should replace this default with something that is little bit smarter. It should factor
in the size of the job and potentially the number of racks available.
> The code that creates the default is [here|https://github.com/apache/incubator-aurora/blob/900a2b299347c71f7c7601a3a3d0a3f0f0afbcb7/src/main/java/org/apache/aurora/scheduler/configuration/ConfigurationManager.java#L142-L153].



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

Mime
View raw message