Feature Request: Make ServiceMonitor API group configurable (support azmonitoring.coreos.com for AKS)

[steeledav]

[rahulait]

@steeledav please provide more info on the feature request

[steeledav]

Thanks Rahul,

Environment
Platform: Azure Kubernetes Service (AKS) v1.35.x
GPU Operator chart version: v25.3.x (deployed via Flux HelmRelease)
Monitoring stack: Azure Managed Prometheus (ama-metrics addon)
Problem
The gpu-operator chart hard-codes monitoring.coreos.com/v1 as the API group for its ServiceMonitor CRD. This is baked into the Go source (resource_manager.go imports promv1 "github.com/prometheus-operator/prometheus-operator/pkg/apis/monitoring/v1") and there is no chart value to override the API group.

On AKS with Azure Managed Prometheus, the ama-metrics agent only scrapes CRDs registered under azmonitoring.coreos.com/v1. ServiceMonitors using monitoring.coreos.com/v1 are ignored entirely — metrics are never collected.

Impact
GPU metrics (DCGM Exporter) are unavailable in Azure Managed Grafana/Prometheus when using the gpu-operator's built-in serviceMonitor.enabled: true
Users must manually create a separate azmonitoring.coreos.com/v1 ServiceMonitor outside the chart as a workaround
This affects all AKS clusters using Azure Managed Prometheus (the default monitoring path for AKS)
Requested Change
Add a configurable value (e.g., serviceMonitor.apiGroup or serviceMonitor.apiVersion) that allows overriding the ServiceMonitor API group from the default monitoring.coreos.com to an alternative like azmonitoring.coreos.com.

Example:
dcgmExporter:  
serviceMonitor:    
enabled: true    
apiGroup: "azmonitoring.coreos.com"  # NEW — defaults to "monitoring.coreos.com"

Workaround
Currently we deploy a separate ServiceMonitor manifest targeting azmonitoring.coreos.com/v1 outside the chart, duplicating configuration that the chart should could natively.