[
https://issues.apache.org/jira/browse/YARN-6620?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16170294#comment-16170294
]
Wangda Tan commented on YARN-6620:
----------------------------------
[~tangzhankun],
Good point, first of all, to me the requirement:
bq. ... one physical machine with two different vendor GPU cards ..
Is not a real requirement, we may need to spend 20% of the effort to support
the 1% requirement.
And even if it is true, as you commented, admin may need to register two
plugins and handle isolation part separately. I can see some minor code changes
needed (For example, make the resource-name inside resource-vector can be
configurable, by default it is gpu, and admin can change it to gpu-vendor-a and
gpu-vendor-b, but I would prefer to do this once the requirement comes.
> [YARN-6223] NM Java side code changes to support isolate GPU devices by using
> CGroups
> -------------------------------------------------------------------------------------
>
> Key: YARN-6620
> URL: https://issues.apache.org/jira/browse/YARN-6620
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Wangda Tan
> Assignee: Wangda Tan
> Attachments: YARN-6620.001.patch, YARN-6620.002.patch,
> YARN-6620.003.patch, YARN-6620.004.patch, YARN-6620.005.patch,
> YARN-6620.006-WIP.patch
>
>
> This JIRA plan to add support of:
> 1) GPU configuration for NodeManagers
> 2) Isolation in CGroups. (Java side).
> 3) NM restart and recovery allocated GPU devices
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
