[
https://issues.apache.org/jira/browse/YUNIKORN-2197?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Wilfred Spiegelenburg resolved YUNIKORN-2197.
---------------------------------------------
Fix Version/s: 1.5.0
Resolution: Fixed
> remove FitInMaxUndef from node code
> -----------------------------------
>
> Key: YUNIKORN-2197
> URL: https://issues.apache.org/jira/browse/YUNIKORN-2197
> Project: Apache YuniKorn
> Issue Type: Bug
> Components: core - scheduler
> Reporter: Wilfred Spiegelenburg
> Assignee: Wilfred Spiegelenburg
> Priority: Critical
> Labels: pull-request-available
> Fix For: 1.5.0
>
>
> With the removal of the "unlimited node" all checks in the node should have
> been reverted to use normal {{FitIn}} checks for resources. If a node does
> not advertise the resource as part of the total resource it should mean none
> are available, and the allocation does not fit.
> The current code still has {{FitInMaxUndef}} in a number of places. This
> would allow an allocation which requests for example 1 GPU to fit on a node
> that does not advertise any GPUs.
> Note that the allocation cycle would not end up allowing the allocation to be
> made as the predicate callback does not pass. It does mean there is extra
> overhead that should be avoided.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
