1. Proposed title of this feature request
Split the TopologyAndDuplicates Descheduler Profile into Separate Topology and Duplicates Profiles

2. What is the nature and description of the request?
The current TopologyAndDuplicates profile in the OpenShift Descheduler Operator combines two eviction strategies:

• Enforcing topology spread constraints

• Evicting duplicate pods on the same node

This request proposes splitting this combined profile into two independent profiles:

• Topology: to handle topology-related evictions (e.g., zone/rack skew)

• Duplicates: to handle evictions based on duplicate pods on the same node

This change would allow customers to disable one strategy (e.g., duplicate eviction) while keeping the other active (e.g., topology enforcement), offering more fine-grained control.

3. Why does the customer need this? (List the business requirements here)

• Avoid unnecessary disruptions: Customers like Suncorp-Metway Ltd report production pods are being unnecessarily evicted due to the duplicate eviction logic, even when topology is already balanced.

• Business continuity: These pod restarts impact production workloads and delay customer transactions.

• Operational flexibility: Customers previously running upstream descheduler had the ability to fine-tune behavior, which is lost with the limited profile granularity in ROSA/ARO.

• Minimal change, high impact: This split is a simple modification but provides significant benefit in terms of configurability and stability.

4. List any affected packages or components.

• cluster-kube-descheduler-operator

• OpenShift Descheduler configuration and CRD definitions

• Profile controller logic within the Descheduler Operator