Added Mtls patch

Bobbins228 · Bobbins228 · commit 3abc7969bd4e · 2024-04-19T11:04:12.000+01:00
(cherry picked from commit de2de96fc88022df783b637ccb145d1d73ba66ff)
diff --git a/config/rbac/role.yaml b/config/rbac/role.yaml
@@ -44,6 +44,12 @@ rules:
   - subjectaccessreviews
   verbs:
   - create
+- apiGroups:
+  - config.openshift.io
+  resources:
+  - ingresses
+  verbs:
+  - get
 - apiGroups:
   - ""
   resources:
diff --git a/main.go b/main.go
@@ -52,6 +52,7 @@ import (
 	"sigs.k8s.io/yaml"
 
 	routev1 "github.com/openshift/api/route/v1"
+	clientset "github.com/openshift/client-go/config/clientset/versioned"
 
 	"github.com/project-codeflare/codeflare-operator/pkg/config"
 	"github.com/project-codeflare/codeflare-operator/pkg/controllers"
@@ -75,6 +76,8 @@ func init() {
 	utilruntime.Must(routev1.Install(scheme))
 }
 
+// +kubebuilder:rbac:groups=config.openshift.io,resources=ingresses,verbs=get;
+
 func main() {
 	var configMapName string
 	flag.StringVar(&configMapName, "config", "codeflare-operator-config",
@@ -117,6 +120,7 @@ func main() {
 		KubeRay: &config.KubeRayConfiguration{
 			RayDashboardOAuthEnabled: ptr.To(true),
 			IngressDomain:            "",
+			MTLSEnabled:              ptr.To(true),
 		},
 	}
 
@@ -155,6 +159,13 @@ func main() {
 	certsReady := make(chan struct{})
 	exitOnError(setupCertManagement(mgr, namespace, certsReady), "unable to setup cert-controller")
 
+	if cfg.KubeRay.IngressDomain == "" {
+		configClient, err := clientset.NewForConfig(kubeConfig)
+		exitOnError(err, "unable to create Route Client Set")
+		cfg.KubeRay.IngressDomain, err = getClusterDomain(ctx, configClient)
+		exitOnError(err, cfg.KubeRay.IngressDomain)
+	}
+
 	go setupControllers(mgr, kubeClient, cfg, isOpenShift(ctx, kubeClient.DiscoveryClient), certsReady)
 
 	setupLog.Info("setting up health endpoints")
@@ -332,3 +343,17 @@ func isOpenShift(ctx context.Context, dc discovery.DiscoveryInterface) bool {
 	logger.Info("We detected being on Vanilla Kubernetes!")
 	return false
 }
+
+func getClusterDomain(ctx context.Context, configClient *clientset.Clientset) (string, error) {
+	ingress, err := configClient.ConfigV1().Ingresses().Get(ctx, "cluster", metav1.GetOptions{})
+	if err != nil {
+		return "", fmt.Errorf("failed to get Ingress object: %v", err)
+	}
+
+	domain := ingress.Spec.Domain
+	if domain == "" {
+		return "", fmt.Errorf("domain is not set in the Ingress object")
+	}
+
+	return domain, nil
+}
diff --git a/pkg/config/config.go b/pkg/config/config.go
@@ -35,6 +35,8 @@ type KubeRayConfiguration struct {
 	RayDashboardOAuthEnabled *bool `json:"rayDashboardOAuthEnabled,omitempty"`
 
 	IngressDomain string `json:"ingressDomain"`
+
+	MTLSEnabled *bool `json:"mTLSEnabled,omitempty"`
 }
 
 type ControllerManager struct {
diff --git a/pkg/controllers/raycluster_webhook.go b/pkg/controllers/raycluster_webhook.go
@@ -18,6 +18,7 @@ package controllers
 
 import (
 	"context"
+	"strconv"
 
 	rayv1 "github.com/ray-project/kuberay/ray-operator/apis/ray/v1"
 
@@ -36,6 +37,7 @@ import (
 const (
 	oauthProxyContainerName = "oauth-proxy"
 	oauthProxyVolumeName    = "proxy-tls-secret"
+	initContainerName       = "create-cert"
 )
 
 // log is for logging in this package.
@@ -66,17 +68,47 @@ var _ webhook.CustomValidator = &rayClusterWebhook{}
 func (w *rayClusterWebhook) Default(ctx context.Context, obj runtime.Object) error {
 	rayCluster := obj.(*rayv1.RayCluster)
 
-	if !ptr.Deref(w.Config.RayDashboardOAuthEnabled, true) {
-		return nil
-	}
+	if ptr.Deref(w.Config.RayDashboardOAuthEnabled, true) {
+		rayclusterlog.V(2).Info("Adding OAuth sidecar container")
+		rayCluster.Spec.HeadGroupSpec.Template.Spec.Containers = upsert(rayCluster.Spec.HeadGroupSpec.Template.Spec.Containers, oauthProxyContainer(rayCluster), withContainerName(oauthProxyContainerName))
 
-	rayclusterlog.V(2).Info("Adding OAuth sidecar container")
+		rayCluster.Spec.HeadGroupSpec.Template.Spec.Volumes = upsert(rayCluster.Spec.HeadGroupSpec.Template.Spec.Volumes, oauthProxyTLSSecretVolume(rayCluster), withVolumeName(oauthProxyVolumeName))
 
-	rayCluster.Spec.HeadGroupSpec.Template.Spec.Containers = upsert(rayCluster.Spec.HeadGroupSpec.Template.Spec.Containers, oauthProxyContainer(rayCluster), withContainerName(oauthProxyContainerName))
+		rayCluster.Spec.HeadGroupSpec.Template.Spec.ServiceAccountName = rayCluster.Name + "-oauth-proxy"
+	}
 
-	rayCluster.Spec.HeadGroupSpec.Template.Spec.Volumes = upsert(rayCluster.Spec.HeadGroupSpec.Template.Spec.Volumes, oauthProxyTLSSecretVolume(rayCluster), withVolumeName(oauthProxyVolumeName))
+	if ptr.Deref(w.Config.MTLSEnabled, true) {
+		rayclusterlog.V(2).Info("Adding create-cert Init Containers")
+		// HeadGroupSpec //
+		// Append the list of environment variables for the ray-head container
+		for index := range rayCluster.Spec.HeadGroupSpec.Template.Spec.Containers {
+			for _, envVar := range envVarList() {
+				rayCluster.Spec.HeadGroupSpec.Template.Spec.Containers[index].Env = upsert(rayCluster.Spec.HeadGroupSpec.Template.Spec.Containers[index].Env, envVar, withEnvVarName(envVar.Name))
+			}
+		}
+
+		// Append the create-cert Init Container
+		rayCluster.Spec.HeadGroupSpec.Template.Spec.InitContainers = upsert(rayCluster.Spec.HeadGroupSpec.Template.Spec.InitContainers, rayHeadInitContainer(rayCluster, w.Config.IngressDomain), withContainerName(initContainerName))
+
+		// Append the CA volumes
+		for _, caVol := range caVolumes(rayCluster) {
+			rayCluster.Spec.HeadGroupSpec.Template.Spec.Volumes = upsert(rayCluster.Spec.HeadGroupSpec.Template.Spec.Volumes, caVol, withVolumeName(caVol.Name))
+		}
+		// WorkerGroupSpec //
+		// Append the list of environment variables for the machine-learning container
+		for index := range rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.Containers {
+			for _, envVar := range envVarList() {
+				rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.Containers[index].Env = upsert(rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.Containers[index].Env, envVar, withEnvVarName(envVar.Name))
+			}
+		}
+		// Append the CA volumes
+		for _, caVol := range caVolumes(rayCluster) {
+			rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.Volumes = upsert(rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.Volumes, caVol, withVolumeName(caVol.Name))
+		}
+		// Append the create-cert Init Container
+		rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.InitContainers = upsert(rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.InitContainers, rayWorkerInitContainer(), withContainerName(initContainerName))
 
-	rayCluster.Spec.HeadGroupSpec.Template.Spec.ServiceAccountName = rayCluster.Name + "-oauth-proxy"
+	}
 
 	return nil
 }
@@ -117,6 +149,14 @@ func (w *rayClusterWebhook) ValidateUpdate(ctx context.Context, oldObj, newObj r
 		allErrors = append(allErrors, validateHeadGroupServiceAccountName(rayCluster)...)
 	}
 
+	// Init Container related errors
+	if ptr.Deref(w.Config.MTLSEnabled, true) {
+		allErrors = append(allErrors, validateHeadInitContainer(rayCluster, w)...)
+		allErrors = append(allErrors, validateWorkerInitContainer(rayCluster)...)
+		allErrors = append(allErrors, validateHeadEnvVars(rayCluster)...)
+		allErrors = append(allErrors, validateWorkerEnvVars(rayCluster)...)
+		allErrors = append(allErrors, validateCaVolumes(rayCluster)...)
+	}
 	return warnings, allErrors.ToAggregate()
 }
 
@@ -225,3 +265,184 @@ func oauthProxyTLSSecretVolume(rayCluster *rayv1.RayCluster) corev1.Volume {
 		},
 	}
 }
+
+func initCaVolumeMounts() []corev1.VolumeMount {
+	return []corev1.VolumeMount{
+		{
+			Name:      "ca-vol",
+			MountPath: "/home/ray/workspace/ca",
+			ReadOnly:  true,
+		},
+		{
+			Name:      "server-cert",
+			MountPath: "/home/ray/workspace/tls",
+			ReadOnly:  false,
+		},
+	}
+}
+
+func envVarList() []corev1.EnvVar {
+	return []corev1.EnvVar{
+		{
+			Name: "MY_POD_IP",
+			ValueFrom: &corev1.EnvVarSource{
+				FieldRef: &corev1.ObjectFieldSelector{
+					FieldPath: "status.podIP",
+				},
+			},
+		},
+		{
+			Name:  "RAY_USE_TLS",
+			Value: "1",
+		},
+		{
+			Name:  "RAY_TLS_SERVER_CERT",
+			Value: "/home/ray/workspace/tls/server.crt",
+		},
+		{
+			Name:  "RAY_TLS_SERVER_KEY",
+			Value: "/home/ray/workspace/tls/server.key",
+		},
+		{
+			Name:  "RAY_TLS_CA_CERT",
+			Value: "/home/ray/workspace/tls/ca.crt",
+		},
+	}
+}
+
+func caVolumes(rayCluster *rayv1.RayCluster) []corev1.Volume {
+	return []corev1.Volume{
+		{
+			Name: "ca-vol",
+			VolumeSource: corev1.VolumeSource{
+				Secret: &corev1.SecretVolumeSource{
+					SecretName: `ca-secret-` + rayCluster.Name,
+				},
+			},
+		},
+		{
+			Name: "server-cert",
+			VolumeSource: corev1.VolumeSource{
+				EmptyDir: &corev1.EmptyDirVolumeSource{},
+			},
+		},
+	}
+}
+
+func rayHeadInitContainer(rayCluster *rayv1.RayCluster, domain string) corev1.Container {
+	rayClientRoute := "rayclient-" + rayCluster.Name + "-" + rayCluster.Namespace + "." + domain
+	// Service name for basic interactive
+	svcDomain := rayCluster.Name + "-head-svc." + rayCluster.Namespace + ".svc"
+
+	initContainerHead := corev1.Container{
+		Name:  "create-cert",
+		Image: "quay.io/project-codeflare/ray:latest-py39-cu118",
+		Command: []string{
+			"sh",
+			"-c",
+			`cd /home/ray/workspace/tls && openssl req -nodes -newkey rsa:2048 -keyout server.key -out server.csr -subj '/CN=ray-head' && printf "authorityKeyIdentifier=keyid,issuer\nbasicConstraints=CA:FALSE\nsubjectAltName = @alt_names\n[alt_names]\nDNS.1 = 127.0.0.1\nDNS.2 = localhost\nDNS.3 = ${FQ_RAY_IP}\nDNS.4 = $(awk 'END{print $1}' /etc/hosts)\nDNS.5 = ` + rayClientRoute + `\nDNS.6 = ` + svcDomain + `">./domain.ext && cp /home/ray/workspace/ca/* . && openssl x509 -req -CA ca.crt -CAkey ca.key -in server.csr -out server.crt -days 365 -CAcreateserial -extfile domain.ext`,
+		},
+		VolumeMounts: initCaVolumeMounts(),
+	}
+	return initContainerHead
+}
+
+func rayWorkerInitContainer() corev1.Container {
+	initContainerWorker := corev1.Container{
+		Name:  "create-cert",
+		Image: "quay.io/project-codeflare/ray:latest-py39-cu118",
+		Command: []string{
+			"sh",
+			"-c",
+			`cd /home/ray/workspace/tls && openssl req -nodes -newkey rsa:2048 -keyout server.key -out server.csr -subj '/CN=ray-head' && printf "authorityKeyIdentifier=keyid,issuer\nbasicConstraints=CA:FALSE\nsubjectAltName = @alt_names\n[alt_names]\nDNS.1 = 127.0.0.1\nDNS.2 = localhost\nDNS.3 = ${FQ_RAY_IP}\nDNS.4 = $(awk 'END{print $1}' /etc/hosts)">./domain.ext && cp /home/ray/workspace/ca/* . && openssl x509 -req -CA ca.crt -CAkey ca.key -in server.csr -out server.crt -days 365 -CAcreateserial -extfile domain.ext`,
+		},
+		VolumeMounts: initCaVolumeMounts(),
+	}
+	return initContainerWorker
+}
+
+func validateHeadInitContainer(rayCluster *rayv1.RayCluster, w *rayClusterWebhook) field.ErrorList {
+	var allErrors field.ErrorList
+
+	if err := contains(rayCluster.Spec.HeadGroupSpec.Template.Spec.InitContainers, rayHeadInitContainer(rayCluster, w.Config.IngressDomain), byContainerName,
+		field.NewPath("spec", "headGroupSpec", "template", "spec", "initContainers"),
+		"create-cert Init Container is immutable"); err != nil {
+		allErrors = append(allErrors, err)
+	}
+
+	return allErrors
+}
+
+func validateWorkerInitContainer(rayCluster *rayv1.RayCluster) field.ErrorList {
+	var allErrors field.ErrorList
+
+	if err := contains(rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.InitContainers, rayWorkerInitContainer(), byContainerName,
+		field.NewPath("spec", "workerGroupSpecs", "0", "template", "spec", "initContainers"),
+		"create-cert Init Container is immutable"); err != nil {
+		allErrors = append(allErrors, err)
+	}
+
+	return allErrors
+}
+
+func validateCaVolumes(rayCluster *rayv1.RayCluster) field.ErrorList {
+	var allErrors field.ErrorList
+
+	for _, caVol := range caVolumes(rayCluster) {
+		if err := contains(rayCluster.Spec.HeadGroupSpec.Template.Spec.Volumes, caVol, byVolumeName,
+			field.NewPath("spec", "headGroupSpec", "template", "spec", "volumes"),
+			"ca-vol and server-cert Secret volumes are immutable"); err != nil {
+			allErrors = append(allErrors, err)
+		}
+		if err := contains(rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.Volumes, caVol, byVolumeName,
+			field.NewPath("spec", "workerGroupSpecs", "0", "template", "spec", "volumes"),
+			"ca-vol and server-cert Secret volumes are immutable"); err != nil {
+			allErrors = append(allErrors, err)
+		}
+	}
+
+	return allErrors
+}
+
+func validateHeadEnvVars(rayCluster *rayv1.RayCluster) field.ErrorList {
+	var allErrors field.ErrorList
+	item := 0
+	for index, container := range rayCluster.Spec.HeadGroupSpec.Template.Spec.Containers {
+		if container.Name == "ray-head" {
+			item = index
+			break
+		}
+	}
+
+	for _, envVar := range envVarList() {
+		if err := contains(rayCluster.Spec.HeadGroupSpec.Template.Spec.Containers[item].Env, envVar, byEnvVarName,
+			field.NewPath("spec", "headGroupSpec", "template", "spec", "containers", strconv.Itoa(item), "env"),
+			"RAY_TLS related environment variables are immutable"); err != nil {
+			allErrors = append(allErrors, err)
+		}
+	}
+
+	return allErrors
+}
+
+func validateWorkerEnvVars(rayCluster *rayv1.RayCluster) field.ErrorList {
+	var allErrors field.ErrorList
+	item := 0
+
+	for index, container := range rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.Containers {
+		if container.Name == "machine-learning" {
+			item = index
+			break
+		}
+	}
+
+	for _, envVar := range envVarList() {
+		if err := contains(rayCluster.Spec.WorkerGroupSpecs[0].Template.Spec.Containers[item].Env, envVar, byEnvVarName,
+			field.NewPath("spec", "workerGroupSpecs", "0", "template", "spec", "containers", strconv.Itoa(item), "env"),
+			"RAY_TLS related environment variables are immutable"); err != nil {
+			allErrors = append(allErrors, err)
+		}
+	}
+
+	return allErrors
+}
diff --git a/pkg/controllers/support.go b/pkg/controllers/support.go
@@ -140,3 +140,14 @@ func withVolumeName(name string) compare[corev1.Volume] {
 		return v1.Name == name
 	}
 }
+
+var byEnvVarName = compare[corev1.EnvVar](
+	func(e1, e2 corev1.EnvVar) bool {
+		return e1.Name == e2.Name
+	})
+
+func withEnvVarName(name string) compare[corev1.EnvVar] {
+	return func(e1, e2 corev1.EnvVar) bool {
+		return e1.Name == name
+	}
+}

Original file line number	Diff line number	Diff line change
`@@ -35,6 +35,8 @@ type KubeRayConfiguration struct {`
`35`	`35`	RayDashboardOAuthEnabled *bool `json:"rayDashboardOAuthEnabled,omitempty"`
`36`	`36`
`37`	`37`	IngressDomain string `json:"ingressDomain"`
	`38`	`+`
	`39`	+ MTLSEnabled *bool `json:"mTLSEnabled,omitempty"`
`38`	`40`	`}`
`39`	`41`
`40`	`42`	`type ControllerManager struct {`