[
https://issues.apache.org/jira/browse/YUNIKORN-1693?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17712668#comment-17712668
]
Craig Condit commented on YUNIKORN-1693:
----------------------------------------
We try to keep the implementation between the standard and plugin deployments
as close as possible. We also have some chicken-and-egg problems with
initialization order that are difficult to overcome (we actually need to
bootstrap some things prior to the first time the plugin handle is provided).
Therefore, I don't think this is a good idea. A second instance of *rest.Config
does not add significant overhead anyway.
> plugin mode should reuse kubeConfig from scheduler framework handle
> -------------------------------------------------------------------
>
> Key: YUNIKORN-1693
> URL: https://issues.apache.org/jira/browse/YUNIKORN-1693
> Project: Apache YuniKorn
> Issue Type: Improvement
> Components: shim - kubernetes
> Reporter: Wei Huang
> Priority: Minor
>
> In the plugin mode, the scheduler framework handle provides a
> {{KubeConfig()}} function to expose a pre-built {{*rest.Config}}, which is
> not used in the current implementation. Instead, shim reads the kubeConfig
> string and constructs {{*rest.Config}} anyways.
> In terms of performance (re-use the same {{*rest.Config}}) and simplifying
> user configuration, can we enhance the current impl. to leverage
> {{hanlde.KubeConfig()}}?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
