You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-issues@hadoop.apache.org by "Sunil Govindan (JIRA)" <ji...@apache.org> on 2018/12/18 05:02:00 UTC

[jira] [Comment Edited] (YARN-9038) [CSI] Add ability to publish/unpublish volumes on node managers

    [ https://issues.apache.org/jira/browse/YARN-9038?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16723696#comment-16723696 ] 

Sunil Govindan edited comment on YARN-9038 at 12/18/18 5:01 AM:
----------------------------------------------------------------

Thanks [~cheersyang]

Few comments:
 # Could u pls add javadoc comments in the protocol, publish/unpublish request/response interfaces for apis.
 # Could we use NM_CSI_DRIVER_PREFIX before NM_CSI_ADAPTOR_PREFIX so that "-adaptor" can be avoided?
 # CsiUtils -> CsiConfigUtils if its only config related management in this class?
 # In proto, if NodePublishVolumeRequest is optional, could that be a last param to ensure all required are in first part?
 # Adding to #2 its tough to understand different b/w csi-driver-adaptor and csi-driver. Could we make it more intuitive?
 # I think at Resource level we should get all the noncountable resource info list. with this, we can improve {{prepareCsiVolumes}}
 # Could we need a DeletionService kind of model to get a successful cleanupCsiVolumes? How stable is cleanup from CSI side?
 # Changes for TestDelegatingLinuxContainerRuntime is not needed.


was (Author: sunilg):
Thanks [~cheersyang]

Few comments:
 # Could u pls add javadoc comments in the protocol, publish/unpublish request/response interfaces for apis.
 # Could we use NM_CSI_DRIVER_PREFIX before NM_CSI_ADAPTOR_PREFIX so that "-adaptor" can be avoided?
 # CsiUtils -> CsiConfigUtils if its only config related management in this class?
 # In proto, if NodePublishVolumeRequest is optional, could that be a last param to ensure all required are in first part?
 # Adding to #2 its tough to understand different b/w csi-driver-adaptor and csi-driver. Could we make it more intuitive?
 # I think at Resource level we should get all the noncountable resource info list. with this, we can improve {{prepareCsiVolumes}}
 # Could we need a DeletionService kind of model to get a successful cleanupCsiVolumes? How stable is cleanup from CSI side?

> [CSI] Add ability to publish/unpublish volumes on node managers
> ---------------------------------------------------------------
>
>                 Key: YARN-9038
>                 URL: https://issues.apache.org/jira/browse/YARN-9038
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>            Reporter: Weiwei Yang
>            Assignee: Weiwei Yang
>            Priority: Major
>              Labels: CSI
>         Attachments: YARN-9038.001.patch, YARN-9038.002.patch
>
>
> We need to add ability to publish volumes on node managers in staging area, under NM's local dir. And then mount the path to docker container to make it visible in the container.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org