You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@yunikorn.apache.org by "Wilfred Spiegelenburg (Jira)" <ji...@apache.org> on 2023/05/05 01:27:00 UTC
[jira] [Commented] (YUNIKORN-1722) excessive resource clone calls in calculateNodesResourceUsage
[ https://issues.apache.org/jira/browse/YUNIKORN-1722?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17719581#comment-17719581 ]
Wilfred Spiegelenburg commented on YUNIKORN-1722:
-------------------------------------------------
This adversely affects the scheduler in a cluster that has a large number of nodes. As per [~pbacsko] comment:
{{nodesResourceUsageMonitor.runOnce()}} runs only once per second and look at the allocations, it's crazy.
!image.png|width=620,height=224!
> excessive resource clone calls in calculateNodesResourceUsage
> -------------------------------------------------------------
>
> Key: YUNIKORN-1722
> URL: https://issues.apache.org/jira/browse/YUNIKORN-1722
> Project: Apache YuniKorn
> Issue Type: Bug
> Components: core - scheduler, metrics
> Reporter: Wilfred Spiegelenburg
> Assignee: Wilfred Spiegelenburg
> Priority: Critical
> Attachments: image.png
>
>
> The node calculateNodesResourceUsage clones the resource object inside a loop. The clone is correct as the allocated resource could be changed while the calculation is happening in the scheduling cycle. However it should not happen inside the loop.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@yunikorn.apache.org
For additional commands, e-mail: issues-help@yunikorn.apache.org