Merge branch 'main' into dependabot/go_modules/k8s.io/api-0.30.3

KPostOffice · web-flow · commit f65ea5d252e7 · 2024-08-21T14:19:36.000-04:00
diff --git a/.github/workflows/e2e_tests.yaml b/.github/workflows/e2e_tests.yaml
@@ -115,7 +115,7 @@ jobs:
 
       - name: Upload logs
         uses: actions/upload-artifact@v4
-        if: always() && steps.deploy.outcome == 'success'
+        if: always() && steps.kind-install.outcome == 'success'
         with:
           name: logs
           retention-days: 10
diff --git a/Makefile b/Makefile
@@ -12,7 +12,7 @@ VERSION ?= v0.0.0-dev
 BUNDLE_VERSION ?= $(VERSION:v%=%)
 
 # APPWRAPPER_VERSION defines the default version of the AppWrapper controller
-APPWRAPPER_VERSION ?= v0.21.1
+APPWRAPPER_VERSION ?= v0.23.0
 APPWRAPPER_REPO ?= github.com/project-codeflare/appwrapper
 APPWRAPPER_CRD ?= ${APPWRAPPER_REPO}/config/crd?ref=${APPWRAPPER_VERSION}
 
diff --git a/README.md b/README.md
@@ -8,9 +8,9 @@ CodeFlare Stack Compatibility Matrix
 
 | Component                    | Version                                                                                           |
 |------------------------------|---------------------------------------------------------------------------------------------------|
-| CodeFlare Operator           | [v1.5.0](https://github.com/project-codeflare/codeflare-operator/releases/tag/v1.5.0)             |
-| CodeFlare-SDK                | [v0.17.0](https://github.com/project-codeflare/codeflare-sdk/releases/tag/v0.17.0)                |
-| AppWrapper                   | [v0.20.2](https://github.com/project-codeflare/appwrapper/releases/tag/v0.20.2)                   |
+| CodeFlare Operator           | [v1.7.0](https://github.com/project-codeflare/codeflare-operator/releases/tag/v1.7.0)             |
+| CodeFlare-SDK                | [v0.19.1](https://github.com/project-codeflare/codeflare-sdk/releases/tag/v0.19.1)                |
+| AppWrapper                   | [v0.23.0](https://github.com/project-codeflare/appwrapper/releases/tag/v0.23.0)                   |
 | KubeRay                      | [v1.1.0](https://github.com/opendatahub-io/kuberay/releases/tag/v1.1.0)                           |
 | Kueue                        | [v0.7.0](https://github.com/opendatahub-io/kueue/releases/tag/v0.7.0)                             |
 <!-- Compatibility Matrix end -->
diff --git a/config/crd/appwrapper/kustomization.yaml b/config/crd/appwrapper/kustomization.yaml
@@ -1,4 +1,4 @@
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
-- github.com/project-codeflare/appwrapper/config/crd?ref=v0.21.1
+- github.com/project-codeflare/appwrapper/config/crd?ref=v0.23.0
diff --git a/config/manager/params.env b/config/manager/params.env
@@ -1,2 +1,2 @@
-codeflare-operator-controller-image=quay.io/opendatahub/codeflare-operator:v1.5.0
+codeflare-operator-controller-image=quay.io/opendatahub/codeflare-operator:v1.7.0
 namespace=opendatahub
diff --git a/config/rbac/role.yaml b/config/rbac/role.yaml
@@ -14,6 +14,14 @@ rules:
   - patch
   - update
   - watch
+- apiGroups:
+  - ""
+  resources:
+  - nodes
+  verbs:
+  - get
+  - list
+  - watch
 - apiGroups:
   - ""
   resources:
diff --git a/go.mod b/go.mod
@@ -3,13 +3,14 @@ module github.com/project-codeflare/codeflare-operator
 go 1.22.2
 
 require (
+	github.com/go-logr/logr v1.4.2
 	github.com/onsi/ginkgo/v2 v2.19.0
 	github.com/onsi/gomega v1.33.1
 	github.com/open-policy-agent/cert-controller v0.10.1
 	github.com/opendatahub-io/opendatahub-operator/v2 v2.10.0
 	github.com/openshift/api v0.0.0-20230823114715-5fdd7511b790
 	github.com/openshift/client-go v0.0.0-20221019143426-16aed247da5c
-	github.com/project-codeflare/appwrapper v0.21.1
+	github.com/project-codeflare/appwrapper v0.23.0
 	github.com/project-codeflare/codeflare-common v0.0.0-20240628111341-56c962a09b7e
 	github.com/ray-project/kuberay/ray-operator v1.1.1
 	go.uber.org/zap v1.27.0
@@ -35,7 +36,7 @@ replace go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp => go.open
 replace github.com/jackc/pgx/v4 => github.com/jackc/pgx/v5 v5.5.4
 
 // These replace directives support the backlevel go version required by ODH build
-replace github.com/project-codeflare/appwrapper v0.21.1 => github.com/project-codeflare/appwrapper v0.21.2-0.20240712173553-5b007c947b37
+replace github.com/project-codeflare/appwrapper v0.23.0 => github.com/project-codeflare/appwrapper v0.23.1-0.20240731154950-ad486fb7e7ee
 
 replace sigs.k8s.io/kueue v0.7.1 => github.com/opendatahub-io/kueue v0.7.0-odh-test
 
@@ -51,7 +52,6 @@ require (
 	github.com/evanphx/json-patch v5.6.0+incompatible // indirect
 	github.com/evanphx/json-patch/v5 v5.8.0 // indirect
 	github.com/fsnotify/fsnotify v1.7.0 // indirect
-	github.com/go-logr/logr v1.4.2 // indirect
 	github.com/go-logr/zapr v1.3.0 // indirect
 	github.com/go-openapi/jsonpointer v0.20.0 // indirect
 	github.com/go-openapi/jsonreference v0.20.2 // indirect
diff --git a/go.sum b/go.sum
@@ -246,8 +246,8 @@ github.com/pkg/errors v0.9.1 h1:FEBLx1zS214owpjy7qsBeixbURkuhQAwrK5UwLGTwt4=
 github.com/pkg/errors v0.9.1/go.mod h1:bwawxfHBFNV+L2hUp1rHADufV3IMtnDRdf1r5NINEl0=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/project-codeflare/appwrapper v0.21.2-0.20240712173553-5b007c947b37 h1:x4qdbN98B9gtaU7pseJWABZzwoDawXLC5QMlx0idXxc=
-github.com/project-codeflare/appwrapper v0.21.2-0.20240712173553-5b007c947b37/go.mod h1:gKjO+iRtMIdBvIBYmN+VciL9kzWmkfwgk/+24wCLhSM=
+github.com/project-codeflare/appwrapper v0.23.1-0.20240731154950-ad486fb7e7ee h1:kl/nuApg9116e1jSmDdUSaNmxG2Q/OpubEqI+YVfizk=
+github.com/project-codeflare/appwrapper v0.23.1-0.20240731154950-ad486fb7e7ee/go.mod h1:6PNty9c0IOonWrJ80j6VL4uTijixtJ3OXSdFJBCDZZE=
 github.com/project-codeflare/codeflare-common v0.0.0-20240628111341-56c962a09b7e h1:juFd1dQyioeMxbVE6F0YD25ozm/jiqJE+MpDhu8p22k=
 github.com/project-codeflare/codeflare-common v0.0.0-20240628111341-56c962a09b7e/go.mod h1:unKTw+XoMANTES3WieG016im7rxZ7IR2/ph++L5Vp1Y=
 github.com/prometheus/client_golang v1.18.0 h1:HzFfmkOzH5Q8L8G+kSJKUx5dtG87sewO+FoDDqP5Tbk=
diff --git a/main.go b/main.go
@@ -116,7 +116,7 @@ func main() {
 	zapOptions.BindFlags(flag.CommandLine)
 	flag.Parse()
 
-	ctrl.SetLogger(zap.New(zap.UseFlagOptions(&zapOptions)))
+	ctrl.SetLogger(controllers.FilteredLogger(zap.New(zap.UseFlagOptions(&zapOptions))))
 	klog.SetLogger(ctrl.Log)
 
 	setupLog.Info("Build info",
diff --git a/pkg/controllers/appwrapper_controller.go b/pkg/controllers/appwrapper_controller.go
@@ -41,3 +41,6 @@ package controllers
 // +kubebuilder:rbac:groups=kueue.x-k8s.io,resources=workloads/finalizers,verbs=update
 // +kubebuilder:rbac:groups=kueue.x-k8s.io,resources=resourceflavors,verbs=get;list;watch
 // +kubebuilder:rbac:groups=kueue.x-k8s.io,resources=workloadpriorityclasses,verbs=get;list;watch
+
+// permission to watch nodes for Autopilot integration
+//+kubebuilder:rbac:groups="",resources=nodes,verbs=get;list;watch
diff --git a/pkg/controllers/raycluster_controller.go b/pkg/controllers/raycluster_controller.go
@@ -213,6 +213,10 @@ func (r *RayClusterReconciler) Reconcile(ctx context.Context, req ctrl.Request)
 			return ctrl.Result{RequeueAfter: requeueTime}, err
 		}
 
+		if err := r.deleteHeadPodIfMissingImagePullSecrets(ctx, cluster); err != nil {
+			return ctrl.Result{RequeueAfter: requeueTime}, err
+		}
+
 		_, err = r.kubeClient.RbacV1().ClusterRoleBindings().Apply(ctx, desiredOAuthClusterRoleBinding(cluster), metav1.ApplyOptions{FieldManager: controllerName, Force: true})
 		if err != nil {
 			logger.Error(err, "Failed to update OAuth ClusterRoleBinding")
@@ -333,9 +337,7 @@ func desiredServiceAccount(cluster *rayv1.RayCluster) *corev1ac.ServiceAccountAp
 				`{"kind":"OAuthRedirectReference","apiVersion":"v1",` +
 				`"reference":{"kind":"Route","name":"` + dashboardNameFromCluster(cluster) + `"}}`,
 		}).
-		WithOwnerReferences(
-			metav1ac.OwnerReference().WithUID(cluster.UID).WithName(cluster.Name).WithKind(cluster.Kind).WithAPIVersion(cluster.APIVersion).WithController(true),
-		)
+		WithOwnerReferences(ownerRefForRayCluster(cluster))
 }
 
 func dashboardNameFromCluster(cluster *rayv1.RayCluster) string {
@@ -357,9 +359,7 @@ func desiredClusterRoute(cluster *rayv1.RayCluster) *routev1ac.RouteApplyConfigu
 				WithTermination(routev1.TLSTerminationReencrypt),
 			),
 		).
-		WithOwnerReferences(
-			metav1ac.OwnerReference().WithUID(cluster.UID).WithName(cluster.Name).WithKind(cluster.Kind).WithAPIVersion(cluster.APIVersion).WithController(true),
-		)
+		WithOwnerReferences(ownerRefForRayCluster(cluster))
 }
 
 func oauthServiceNameFromCluster(cluster *rayv1.RayCluster) string {
@@ -385,9 +385,7 @@ func desiredOAuthService(cluster *rayv1.RayCluster) *corev1ac.ServiceApplyConfig
 				).
 				WithSelector(map[string]string{"ray.io/cluster": cluster.Name, "ray.io/node-type": "head"}),
 		).
-		WithOwnerReferences(
-			metav1ac.OwnerReference().WithUID(cluster.UID).WithName(cluster.Name).WithKind(cluster.Kind).WithAPIVersion(cluster.APIVersion).WithController(true),
-		)
+		WithOwnerReferences(ownerRefForRayCluster(cluster))
 }
 
 func oauthSecretNameFromCluster(cluster *rayv1.RayCluster) string {
@@ -404,9 +402,7 @@ func desiredOAuthSecret(cluster *rayv1.RayCluster, cookieSalt string) *corev1ac.
 	return corev1ac.Secret(oauthSecretNameFromCluster(cluster), cluster.Namespace).
 		WithLabels(map[string]string{RayClusterNameLabel: cluster.Name}).
 		WithStringData(map[string]string{"cookie_secret": cookieSecret}).
-		WithOwnerReferences(
-			metav1ac.OwnerReference().WithUID(cluster.UID).WithName(cluster.Name).WithKind(cluster.Kind).WithAPIVersion(cluster.APIVersion).WithController(true),
-		)
+		WithOwnerReferences(ownerRefForRayCluster(cluster))
 }
 
 func caSecretNameFromCluster(cluster *rayv1.RayCluster) string {
@@ -420,12 +416,7 @@ func desiredCASecret(cluster *rayv1.RayCluster, key, cert []byte) *corev1ac.Secr
 			CAPrivateKeyKey: key,
 			CACertKey:       cert,
 		}).
-		WithOwnerReferences(metav1ac.OwnerReference().
-			WithUID(cluster.UID).
-			WithName(cluster.Name).
-			WithKind(cluster.Kind).
-			WithAPIVersion(cluster.APIVersion).
-			WithController(true))
+		WithOwnerReferences(ownerRefForRayCluster(cluster))
 }
 
 func generateCACertificate() ([]byte, []byte, error) {
@@ -470,6 +461,7 @@ func generateCACertificate() ([]byte, []byte, error) {
 
 	return privateKeyPem, certPem, nil
 }
+
 func desiredWorkersNetworkPolicy(cluster *rayv1.RayCluster) *networkingv1ac.NetworkPolicyApplyConfiguration {
 	return networkingv1ac.NetworkPolicy(cluster.Name+"-workers", cluster.Namespace).
 		WithLabels(map[string]string{RayClusterNameLabel: cluster.Name}).
@@ -482,10 +474,9 @@ func desiredWorkersNetworkPolicy(cluster *rayv1.RayCluster) *networkingv1ac.Netw
 					),
 			),
 		).
-		WithOwnerReferences(
-			metav1ac.OwnerReference().WithUID(cluster.UID).WithName(cluster.Name).WithKind(cluster.Kind).WithAPIVersion(cluster.APIVersion).WithController(true),
-		)
+		WithOwnerReferences(ownerRefForRayCluster(cluster))
 }
+
 func desiredHeadNetworkPolicy(cluster *rayv1.RayCluster, cfg *config.KubeRayConfiguration, kubeRayNamespaces []string) *networkingv1ac.NetworkPolicyApplyConfiguration {
 	allSecuredPorts := []*networkingv1ac.NetworkPolicyPortApplyConfiguration{
 		networkingv1ac.NetworkPolicyPort().WithProtocol(corev1.ProtocolTCP).WithPort(intstr.FromInt(8443)),
@@ -539,9 +530,50 @@ func desiredHeadNetworkPolicy(cluster *rayv1.RayCluster, cfg *config.KubeRayConf
 					),
 			),
 		).
-		WithOwnerReferences(
-			metav1ac.OwnerReference().WithUID(cluster.UID).WithName(cluster.Name).WithKind(cluster.Kind).WithAPIVersion(cluster.APIVersion).WithController(true),
-		)
+		WithOwnerReferences(ownerRefForRayCluster(cluster))
+}
+
+func (r *RayClusterReconciler) deleteHeadPodIfMissingImagePullSecrets(ctx context.Context, cluster *rayv1.RayCluster) error {
+	serviceAccount, err := r.kubeClient.CoreV1().ServiceAccounts(cluster.Namespace).Get(ctx, oauthServiceAccountNameFromCluster(cluster), metav1.GetOptions{})
+	if err != nil {
+		return fmt.Errorf("failed to get OAuth ServiceAccount: %w", err)
+	}
+
+	headPod, err := getHeadPod(ctx, r, cluster)
+	if err != nil {
+		return fmt.Errorf("failed to get head pod: %w", err)
+	}
+
+	if headPod == nil {
+		return nil
+	}
+
+	missingSecrets := map[string]bool{}
+	for _, secret := range serviceAccount.ImagePullSecrets {
+		missingSecrets[secret.Name] = true
+	}
+	for _, secret := range headPod.Spec.ImagePullSecrets {
+		delete(missingSecrets, secret.Name)
+	}
+	if len(missingSecrets) > 0 {
+		if err := r.kubeClient.CoreV1().Pods(headPod.Namespace).Delete(ctx, headPod.Name, metav1.DeleteOptions{}); err != nil {
+			return fmt.Errorf("failed to delete head pod: %w", err)
+		}
+	}
+	return nil
+}
+
+func getHeadPod(ctx context.Context, r *RayClusterReconciler, cluster *rayv1.RayCluster) (*corev1.Pod, error) {
+	podList, err := r.kubeClient.CoreV1().Pods(cluster.Namespace).List(ctx, metav1.ListOptions{
+		LabelSelector: fmt.Sprintf("ray.io/node-type=head,ray.io/cluster=%s", cluster.Name),
+	})
+	if err != nil {
+		return nil, err
+	}
+	if len(podList.Items) > 0 {
+		return &podList.Items[0], nil
+	}
+	return nil, nil
 }
 
 // SetupWithManager sets up the controller with the Manager.
@@ -577,7 +609,8 @@ func (r *RayClusterReconciler) SetupWithManager(mgr ctrl.Manager) error {
 					NamespacedName: client.ObjectKey{
 						Name:      name,
 						Namespace: namespace,
-					}}}
+					},
+				}}
 			}),
 		)
 	if r.IsOpenShift {
diff --git a/pkg/controllers/raycluster_controller_test.go b/pkg/controllers/raycluster_controller_test.go
@@ -33,7 +33,7 @@ import (
 
 var _ = Describe("RayCluster controller", func() {
 	Context("RayCluster controller test", func() {
-		var rayClusterName = "test-raycluster"
+		rayClusterName := "test-raycluster"
 		var namespaceName string
 		BeforeEach(func(ctx SpecContext) {
 			By("Creating a namespace for running the tests.")
@@ -145,6 +145,53 @@ var _ = Describe("RayCluster controller", func() {
 			}).WithTimeout(time.Second * 10).Should(WithTransform(OwnerReferenceName, Equal(foundRayCluster.Name)))
 		})
 
+		It("should delete the head pod if missing image pull secrets", func(ctx SpecContext) {
+			foundRayCluster, err := rayClient.RayV1().RayClusters(namespaceName).Get(ctx, rayClusterName, metav1.GetOptions{})
+			Expect(err).To(Not(HaveOccurred()))
+
+			Eventually(func() (*corev1.ServiceAccount, error) {
+				return k8sClient.CoreV1().ServiceAccounts(namespaceName).Get(ctx, oauthServiceAccountNameFromCluster(foundRayCluster), metav1.GetOptions{})
+			}).WithTimeout(time.Second * 10).Should(WithTransform(OwnerReferenceKind, Equal("RayCluster")))
+
+			headPodName := "head-pod"
+			headPod := &corev1.Pod{
+				ObjectMeta: metav1.ObjectMeta{
+					Name:      headPodName,
+					Namespace: namespaceName,
+					Labels: map[string]string{
+						"ray.io/node-type": "head",
+						"ray.io/cluster":   foundRayCluster.Name,
+					},
+				},
+				Spec: corev1.PodSpec{
+					Containers: []corev1.Container{
+						{
+							Name:  "head-container",
+							Image: "busybox",
+						},
+					},
+				},
+			}
+			_, err = k8sClient.CoreV1().Pods(namespaceName).Create(ctx, headPod, metav1.CreateOptions{})
+			Expect(err).To(Not(HaveOccurred()))
+
+			Eventually(func() (*corev1.Pod, error) {
+				return k8sClient.CoreV1().Pods(namespaceName).Get(ctx, headPodName, metav1.GetOptions{})
+			}).WithTimeout(time.Second * 10).ShouldNot(BeNil())
+
+			sa, err := k8sClient.CoreV1().ServiceAccounts(namespaceName).Get(ctx, oauthServiceAccountNameFromCluster(foundRayCluster), metav1.GetOptions{})
+			Expect(err).To(Not(HaveOccurred()))
+
+			sa.ImagePullSecrets = append(sa.ImagePullSecrets, corev1.LocalObjectReference{Name: "test-image-pull-secret"})
+			_, err = k8sClient.CoreV1().ServiceAccounts(namespaceName).Update(ctx, sa, metav1.UpdateOptions{})
+			Expect(err).To(Not(HaveOccurred()))
+
+			Eventually(func() error {
+				_, err := k8sClient.CoreV1().Pods(namespaceName).Get(ctx, headPodName, metav1.GetOptions{})
+				return err
+			}).WithTimeout(time.Second * 10).Should(Satisfy(errors.IsNotFound))
+		})
+
 		It("should remove CRB when the RayCluster is deleted", func(ctx SpecContext) {
 			foundRayCluster, err := rayClient.RayV1().RayClusters(namespaceName).Get(ctx, rayClusterName, metav1.GetOptions{})
 			Expect(err).To(Not(HaveOccurred()))
@@ -157,7 +204,6 @@ var _ = Describe("RayCluster controller", func() {
 				return err
 			}).WithTimeout(time.Second * 10).Should(Satisfy(errors.IsNotFound))
 		})
-
 	})
 })
 
diff --git a/pkg/controllers/raycluster_webhook_test.go b/pkg/controllers/raycluster_webhook_test.go
diff --git a/pkg/controllers/support.go b/pkg/controllers/support.go

Original file line number	Diff line number	Diff line change
`@@ -1,2 +1,2 @@`
`1`		`-codeflare-operator-controller-image=quay.io/opendatahub/codeflare-operator:v1.5.0`
	`1`	`+codeflare-operator-controller-image=quay.io/opendatahub/codeflare-operator:v1.7.0`
`2`	`2`	`namespace=opendatahub`