[
https://issues.apache.org/jira/browse/YUNIKORN-2279?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Wilfred Spiegelenburg updated YUNIKORN-2279:
--------------------------------------------
Parent: YUNIKORN-3186
Issue Type: Sub-task (was: Improvement)
> requiredNodePreemptor may preempt ask acquired resources through normal
> preemption
> ----------------------------------------------------------------------------------
>
> Key: YUNIKORN-2279
> URL: https://issues.apache.org/jira/browse/YUNIKORN-2279
> Project: Apache YuniKorn
> Issue Type: Sub-task
> Components: core - scheduler
> Reporter: Manikandan R
> Assignee: Manikandan R
> Priority: Major
>
> A ask (preemptor) acquired resources through normal preemption process may be
> preempted by requiredNodePreemptor to free up spaces for daemon sets. It may
> happen once ask (preemptor) starts running or reserved on the node. So, whole
> preemption effort in making up spaces for this ask (preemptor) won't be
> beneficial as expected.
> Currently requiredNodePreemptor skip only daemon sets while choosing victims
> and can cancel any reservations made on the node to acquire resources. So, we
> need to differentiate the "normal" reservations from "preempted" reservations
> to help the requiredNode preemptor to choose the victims. In addition, node
> having "preempted" reservations can be skipped for normal preemption process
> because node already has preemptor candidate to run.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
