You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@yunikorn.apache.org by "Weiwei Yang (Jira)" <ji...@apache.org> on 2020/06/29 23:44:00 UTC

[jira] [Comment Edited] (YUNIKORN-260) Handle node events in shim side

    [ https://issues.apache.org/jira/browse/YUNIKORN-260?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17148215#comment-17148215 ] 

Weiwei Yang edited comment on YUNIKORN-260 at 6/29/20, 11:43 PM:
-----------------------------------------------------------------

Hi [~adam.antal]

I think we need to postpone this to post 0.9 release. I have a couple of concerns:
 * Are we publishing events from the core now? 
 * the PR in #146 will cause the shim to query API server once the {{PublishEvents}} is called. this could be adding a big overhead to api-server. Instead of doing this, we should leverage the cached nodes in {{pkg/cache/external/scheduler_cache.go}}. we should be able to get the node object by ctx.schedulerCache.GetNode(node).Node()


was (Author: wwei):
Hi [~adam.antal]

I think we need to postpone this to post 0.9 release. I have a couple of concerns:
 * Are we publishing events from the core now? 
 * the PR in #146 will cause the shim to query API server once the \{{PublishEvents}} is called. this could be adding a big overhead to api-server. Instead of doing this, we should leverage the cached nodes in \{{pkg/cache/external/scheduler_cache.go}}. we should be able to get the node object by \{{ctx.schedulerCache.GetNode(node).Node()}}

> Handle node events in shim side
> -------------------------------
>
>                 Key: YUNIKORN-260
>                 URL: https://issues.apache.org/jira/browse/YUNIKORN-260
>             Project: Apache YuniKorn
>          Issue Type: Sub-task
>          Components: shim - kubernetes
>            Reporter: Adam Antal
>            Assignee: Adam Antal
>            Priority: Major
>              Labels: pull-request-available
>
> Right now we only emit events for allocations (pods), but we have multiple types of events. Let's add handling the node events as they are fairly simple.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

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