hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Wangda Tan (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (YARN-4026) FiCaSchedulerApp: Revisit ContainerAllocator interface design
Date Fri, 07 Aug 2015 05:36:45 GMT

     [ https://issues.apache.org/jira/browse/YARN-4026?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Wangda Tan updated YARN-4026:
-----------------------------
    Description: 
After YARN-3983, we have an extensible ContainerAllocator which can be used by FiCaSchedulerApp
to decide how to allocate resources.

While working on YARN-1651 (allocate resource to increase container), I found one thing in
existing logic not flexible enough:
- ContainerAllocator decides what to allocate for a given node and priority: To support different
kinds of resource allocation, for example, priority as weight / skip priority or not, etc.
It's better to let ContainerAllocator to choose how to order resource request according to
priorities.


  was:
After YARN-3983, we have an extensible ContainerAllocator which can be used by FiCaSchedulerApp
to decide how to allocate resources.

While working on YARN-1651 (allocate resource to increase container), I found existing logic:
- ContainerAllocator decides what to allocate for a given node and priority, it's not flexible
enough: To support different kinds of resource allocation, for example, priority as weight
/ skip priority or not, etc. It's better to let ContainerAllocator to choose how to order
resource request according to priorities.



> FiCaSchedulerApp: Revisit ContainerAllocator interface design
> -------------------------------------------------------------
>
>                 Key: YARN-4026
>                 URL: https://issues.apache.org/jira/browse/YARN-4026
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Wangda Tan
>            Assignee: Wangda Tan
>
> After YARN-3983, we have an extensible ContainerAllocator which can be used by FiCaSchedulerApp
to decide how to allocate resources.
> While working on YARN-1651 (allocate resource to increase container), I found one thing
in existing logic not flexible enough:
> - ContainerAllocator decides what to allocate for a given node and priority: To support
different kinds of resource allocation, for example, priority as weight / skip priority or
not, etc. It's better to let ContainerAllocator to choose how to order resource request according
to priorities.



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

Mime
View raw message