XML

Word

Printable

Epic Name:
ga-cpu-load-aware-rebalancing-descheduler
Activity Type:
Product / Portfolio Work
Story Points:
77
Current Status:
Green
Epic Status:
In Progress
Feature Link:
VIRTSTRAT-65 - CPU Load Aware balancing within a single cluster
Parent Link:
VIRTSTRAT-65CPU Load Aware balancing within a single cluster
Hierarchy Progress Bar:

11% To Do, 44% In Progress, 44% Done
Status Summary:

Hide

2025-10-06:
Dashboard got merged into ACM. OCPBUGS-62301 is related to this feature, but currently not a blocker....

Show
2025-10-06: Dashboard got merged into ACM. OCPBUGS-62301 is related to this feature, but currently not a blocker....

Goal

This is a follow up from epic ~~CNV-51099~~ to be able to ship our load aware balancing solution as GA.
Remaining work includes:

a dashboard to help the cluster admin monitor load aware balancing
cluster admin visible alerts when the descheduler or its dependencies are not properly configured
a mechanism to let a VM owner opt-in/opt-out for the single VM without the need to reboot it

User Stories

As a cluster administrator, I want have a dedicated dashboard to monitor load aware balancing, so that I can monitor if my cluster is balanced and how the load aware balancing is working
As a cluster administrator, I want to receive an alert if load aware balancing is not properly configured for my cluster
As a VM owner, I want have an opt-in/opt-out mechanism for my VM, so that I can can easily opt-in/opt-out without the need to reboot my VM (setting an higher priority is also an option but it currently requires a reboot)

is blocked by

CNV-67105 eviction annotation is signaling success despite eviction failed (or in progress)

is related to

CNV-66703 [UI] use a new annotation to eventually opt-out selected VMs from the the descheduler scope

relates to

CNV-67925 Eviction can last forever for VMs that fails to be migrated

OCPBUGS-62301 Evaluation of platform default kernel psi argument impact and Kube Descheduler Guidance

split from

CNV-51099 TP: CPU Load Aware rebalancing with Descheduler

1.	upstream roadmap issue	Closed	Unassigned
2.	upstream design	Closed	Unassigned
3.	upstream documentation	Closed	Unassigned
4.	upgrade consideration	Closed	Unassigned
5.	CEE/PX summary presentation	Closed	Unassigned
6.	test plans in polarion	Closed	Denys Shchedrivyi
7.	automated tests	Closed	Denys Shchedrivyi
8.	downstream documentation merged	New	Unassigned