Fix issue with head pod not monitered by Prometheus under certain condition

[Yicheng-Lu-llll]

Why are these changes needed?

KubeRay is anticipated to provide a Prometheus metrics endpoint on port 8080.

However, if users specify alternative ports in Spec.HeadGroupSpec.Template.Spec.Containers[ray_index].Ports without setting the metrics port, the head service will exclude the default metrics port. As a result, the service monitor will be unable to scrape metrics. Refer to the following code snippet for the existing logic.

kuberay/ray-operator/controllers/ray/common/service.go

Lines 158 to 163 in 6490749

	func getServicePorts(cluster rayiov1alpha1.RayCluster) map[string]int32 {
	ports, err := getPortsFromCluster(cluster)
	// Assign default ports
	if err != nil || len(ports) == 0 {
	ports = getDefaultPorts()
	}

This can be reproduced by using any sample yaml files, for example, running:

kind delete cluster
kind create cluster
# kuberay repo path
rootPath="/home/ubuntu/kuberay" 
cd $rootPath
# install the kube-prometheus-stack chart and related custom resources, including ServiceMonitor, PodMonitor and PrometheusRule
./install/prometheus/install.sh  

# install kuberay-operator
helm install kuberay-operator kuberay/kuberay-operator --version 0.4.0

# install ray cluster
kubectl apply -f $rootPath/ray-operator/config/samples/ray-cluster.complete.yaml

kubectl port-forward --address 0.0.0.0 prometheus-prometheus-kube-prometheus-prometheus-0 -n prometheus-system 9090:9090
# see result in http://localhost:9090/targets?search=ray. Only ray-workers-monitor is set.

Related issue number

Checks

After this PR, ray-workers-monitor and ray-head-monitor are all set.

[kevin85421]

Would you mind adding a test?

[kevin85421]

LGTM