Merge pull request #135 from mshitrit/same-kind-support

Add support to multiple templates with the same kind

PROJECT

@@ -33,6 +33,7 @@ resources:
   path: github.com/medik8s/fence-agents-remediation/api/v1alpha1
   version: v1alpha1
   webhooks:
+    defaulting: true
     validation: true
     webhookVersion: v1
 version: "3"

api/v1alpha1/fenceagentsremediation_webhook.go

@@ -47,6 +47,7 @@ func (r *FenceAgentsRemediation) SetupWebhookWithManager(mgr ctrl.Manager) error
 
 // TODO(user): change verbs to "verbs=create;update;delete" if you want to enable deletion validation.
 // +kubebuilder:webhook:path=/validate-fence-agents-remediation-medik8s-io-v1alpha1-fenceagentsremediation,mutating=false,failurePolicy=fail,sideEffects=None,groups=fence-agents-remediation.medik8s.io,resources=fenceagentsremediations,verbs=create;update,versions=v1alpha1,name=vfenceagentsremediation.kb.io,admissionReviewVersions=v1
+
 var _ webhook.Validator = &FenceAgentsRemediation{}
 
 // ValidateCreate implements webhook.Validator so a webhook will be registered for the type

api/v1alpha1/fenceagentsremediationtemplate_webhook.go

@@ -17,6 +17,8 @@ limitations under the License.
 package v1alpha1
 
 import (
+	commonAnnotations "github.com/medik8s/common/pkg/annotations"
+
 	"k8s.io/apimachinery/pkg/runtime"
 	ctrl "sigs.k8s.io/controller-runtime"
 	logf "sigs.k8s.io/controller-runtime/pkg/log"
@@ -37,8 +39,24 @@ func (r *FenceAgentsRemediationTemplate) SetupWebhookWithManager(mgr ctrl.Manage
 
 // TODO(user): EDIT THIS FILE!  THIS IS SCAFFOLDING FOR YOU TO OWN!
 
+// +kubebuilder:webhook:path=/mutate-fence-agents-remediation-medik8s-io-v1alpha1-fenceagentsremediationtemplate,mutating=true,failurePolicy=fail,sideEffects=None,groups=fence-agents-remediation.medik8s.io,resources=fenceagentsremediationtemplates,verbs=create;update,versions=v1alpha1,name=mfenceagentsremediationtemplate.kb.io,admissionReviewVersions=v1
+
+var _ webhook.Defaulter = &FenceAgentsRemediationTemplate{}
+
+// Default implements webhook.Defaulter so a webhook will be registered for the type
+func (farTemplate *FenceAgentsRemediationTemplate) Default() {
+	webhookFARTemplateLog.Info("default", "name", farTemplate.Name)
+	if farTemplate.GetAnnotations() == nil {
+		farTemplate.Annotations = make(map[string]string)
+	}
+	if _, isSameKindAnnotationSet := farTemplate.GetAnnotations()[commonAnnotations.MultipleTemplatesSupportedAnnotation]; !isSameKindAnnotationSet {
+		farTemplate.Annotations[commonAnnotations.MultipleTemplatesSupportedAnnotation] = "true"
+	}
+}
+
 // TODO(user): change verbs to "verbs=create;update;delete" if you want to enable deletion validation.
 // +kubebuilder:webhook:path=/validate-fence-agents-remediation-medik8s-io-v1alpha1-fenceagentsremediationtemplate,mutating=false,failurePolicy=fail,sideEffects=None,groups=fence-agents-remediation.medik8s.io,resources=fenceagentsremediationtemplates,verbs=create;update,versions=v1alpha1,name=vfenceagentsremediationtemplate.kb.io,admissionReviewVersions=v1
+
 var _ webhook.Validator = &FenceAgentsRemediationTemplate{}
 
 // ValidateCreate implements webhook.Validator so a webhook will be registered for the type

bundle/manifests/fence-agents-remediation.clusterserviceversion.yaml

@@ -410,6 +410,26 @@ spec:
   replaces: fence-agents-remediation.v0.0.1
   version: 0.0.1
   webhookdefinitions:
+  - admissionReviewVersions:
+    - v1
+    containerPort: 443
+    deploymentName: fence-agents-remediation-controller-manager
+    failurePolicy: Fail
+    generateName: mfenceagentsremediationtemplate.kb.io
+    rules:
+    - apiGroups:
+      - fence-agents-remediation.medik8s.io
+      apiVersions:
+      - v1alpha1
+      operations:
+      - CREATE
+      - UPDATE
+      resources:
+      - fenceagentsremediationtemplates
+    sideEffects: None
+    targetPort: 9443
+    type: MutatingAdmissionWebhook
+    webhookPath: /mutate-fence-agents-remediation-medik8s-io-v1alpha1-fenceagentsremediationtemplate
   - admissionReviewVersions:
     - v1
     containerPort: 443

config/webhook/manifests.yaml

@@ -1,5 +1,31 @@
 ---
 apiVersion: admissionregistration.k8s.io/v1
+kind: MutatingWebhookConfiguration
+metadata:
+  name: mutating-webhook-configuration
+webhooks:
+- admissionReviewVersions:
+  - v1
+  clientConfig:
+    service:
+      name: webhook-service
+      namespace: system
+      path: /mutate-fence-agents-remediation-medik8s-io-v1alpha1-fenceagentsremediationtemplate
+  failurePolicy: Fail
+  name: mfenceagentsremediationtemplate.kb.io
+  rules:
+  - apiGroups:
+    - fence-agents-remediation.medik8s.io
+    apiVersions:
+    - v1alpha1
+    operations:
+    - CREATE
+    - UPDATE
+    resources:
+    - fenceagentsremediationtemplates
+  sideEffects: None
+---
+apiVersion: admissionregistration.k8s.io/v1
 kind: ValidatingWebhookConfiguration
 metadata:
   name: validating-webhook-configuration

controllers/fenceagentsremediation_controller.go

@@ -120,13 +120,13 @@ func (r *FenceAgentsRemediationReconciler) Reconcile(ctx context.Context, req ct
 
 	// Validate FAR CR name to match a nodeName from the cluster
 	r.Log.Info("Check FAR CR's name")
-	node, err := utils.GetNodeWithName(r.Client, req.Name)
+	node, err := utils.GetNodeWithName(r.Client, getNodeName(far))
 	if err != nil {
 		r.Log.Error(err, "Unexpected error when validating CR's name with nodes' names", "CR's Name", req.Name)
 		return emptyResult, err
 	}
 	if node == nil {
-		r.Log.Error(err, "Didn't find a node matching the CR's name", "CR's Name", req.Name)
+		r.Log.Error(err, "Could not find CR's target node", "CR's Name", req.Name, "Expected node name", getNodeName(far))
 		utils.UpdateConditions(utils.RemediationFinishedNodeNotFound, far, r.Log)
 		commonEvents.WarningEvent(r.Recorder, far, utils.EventReasonCrNodeNotFound, utils.EventMessageCrNodeNotFound)
 		return emptyResult, err
@@ -168,7 +168,7 @@ func (r *FenceAgentsRemediationReconciler) Reconcile(ctx context.Context, req ct
 
 		// remove node's taints
 		taint := utils.CreateRemediationTaint()
-		if err := utils.RemoveTaint(r.Client, far.Name, taint); err != nil {
+		if err := utils.RemoveTaint(r.Client, node.Name, taint); err != nil {
 			if apiErrors.IsConflict(err) {
 				r.Log.Info("Failed to remove taint from node due to node update, retrying... ,", "node name", node.Name, "taint key", taint.Key, "taint effect", taint.Effect)
 				return ctrl.Result{RequeueAfter: time.Second}, nil
@@ -179,7 +179,7 @@ func (r *FenceAgentsRemediationReconciler) Reconcile(ctx context.Context, req ct
 			}
 		}
 
-		r.Log.Info("FAR remediation taint was removed", "Node Name", req.Name)
+		r.Log.Info("FAR remediation taint was removed", "Node Name", node.Name)
 		commonEvents.NormalEvent(r.Recorder, node, utils.EventReasonRemoveRemediationTaint, utils.EventMessageRemoveRemediationTaint)
 		// remove finalizer
 		controllerutil.RemoveFinalizer(far, v1alpha1.FARFinalizer)
@@ -191,11 +191,11 @@ func (r *FenceAgentsRemediationReconciler) Reconcile(ctx context.Context, req ct
 		return emptyResult, nil
 	}
 	// Add FAR (medik8s) remediation taint
-	taintAdded, err := utils.AppendTaint(r.Client, far.Name)
+	taintAdded, err := utils.AppendTaint(r.Client, node.Name)
 	if err != nil {
 		return emptyResult, err
 	} else if taintAdded {
-		r.Log.Info("FAR remediation taint was added", "Node Name", req.Name)
+		r.Log.Info("FAR remediation taint was added", "Node Name", node.Name)
 		commonEvents.NormalEvent(r.Recorder, node, utils.EventReasonAddRemediationTaint, utils.EventMessageAddRemediationTaint)
 	}
 
@@ -204,19 +204,19 @@ func (r *FenceAgentsRemediationReconciler) Reconcile(ctx context.Context, req ct
 		// The remeditation has already been processed, thus we can begin with executing the FA for the node
 
 		if r.Executor.Exists(far.GetUID()) {
-			r.Log.Info("A Fence Agent is already running", "Fence Agent", far.Spec.Agent, "Node Name", req.Name, "FAR uid", far.GetUID())
+			r.Log.Info("A Fence Agent is already running", "Fence Agent", far.Spec.Agent, "Node Name", node.Name, "FAR uid", far.GetUID())
 			return emptyResult, nil
 		}
 
-		r.Log.Info("Build fence agent command line", "Fence Agent", far.Spec.Agent, "Node Name", req.Name)
+		r.Log.Info("Build fence agent command line", "Fence Agent", far.Spec.Agent, "Node Name", node.Name)
 		faParams, err := buildFenceAgentParams(far)
 		if err != nil {
-			r.Log.Error(err, "Invalid shared or node parameters from CR", "Name", req.Name)
+			r.Log.Error(err, "Invalid shared or node parameters from CR", "Node Name", node.Name, "CR Name", req.Name)
 			return emptyResult, nil
 		}
 
 		cmd := append([]string{far.Spec.Agent}, faParams...)
-		r.Log.Info("Execute the fence agent", "Fence Agent", far.Spec.Agent, "Node Name", req.Name, "FAR uid", far.GetUID())
+		r.Log.Info("Execute the fence agent", "Fence Agent", far.Spec.Agent, "Node Name", node.Name, "FAR uid", far.GetUID())
 		r.Executor.AsyncExecute(ctx, far.GetUID(), cmd, far.Spec.RetryCount, far.Spec.RetryInterval.Duration, far.Spec.Timeout.Duration)
 		commonEvents.NormalEvent(r.Recorder, far, utils.EventReasonFenceAgentExecuted, utils.EventMessageFenceAgentExecuted)
 		return emptyResult, nil
@@ -228,16 +228,16 @@ func (r *FenceAgentsRemediationReconciler) Reconcile(ctx context.Context, req ct
 		// - try to remove workloads
 		// - clean up Executor routine
 
-		r.Log.Info("Manual workload deletion", "Node Name", req.Name)
+		r.Log.Info("Manual workload deletion", "Node Name", node.Name)
 		commonEvents.NormalEvent(r.Recorder, node, utils.EventReasonDeleteResources, utils.EventMessageDeleteResources)
-		if err := commonResources.DeletePods(ctx, r.Client, req.Name); err != nil {
+		if err := commonResources.DeletePods(ctx, r.Client, node.Name); err != nil {
 			r.Log.Error(err, "Manual workload deletion has failed", "CR's Name", req.Name)
 			return emptyResult, err
 		}
 		utils.UpdateConditions(utils.RemediationFinishedSuccessfully, far, r.Log)
 
 		r.Executor.Remove(far.GetUID())
-		r.Log.Info("FenceAgentsRemediation CR has completed to remediate the node", "Node Name", req.Name)
+		r.Log.Info("FenceAgentsRemediation CR has completed to remediate the node", "Node Name", node.Name)
 		commonEvents.NormalEvent(r.Recorder, node, utils.EventReasonNodeRemediationCompleted, utils.EventMessageNodeRemediationCompleted)
 		commonEvents.RemediationFinished(r.Recorder, far)
 	}
@@ -287,6 +287,18 @@ func (r *FenceAgentsRemediationReconciler) updateStatus(ctx context.Context, far
 	return nil
 }
 
+// getNodeName checks for the node name in far's commonAnnotations.NodeNameAnnotation if it does not exist it assumes the node name equals to far CR's name and return it.
+func getNodeName(far *v1alpha1.FenceAgentsRemediation) string {
+	ann := far.GetAnnotations()
+	if ann == nil {
+		return far.GetName()
+	}
+	if nodeName, isNodeNameAnnotationExist := ann[commonAnnotations.NodeNameAnnotation]; isNodeNameAnnotationExist {
+		return nodeName
+	}
+	return far.GetName()
+}
+
 // buildFenceAgentParams collects the FAR's parameters for the node based on FAR CR, and if the CR is missing parameters
 // or the CR's name don't match nodeParameter name or it has an action which is different than reboot, then return an error
 func buildFenceAgentParams(far *v1alpha1.FenceAgentsRemediation) ([]string, error) {
@@ -314,7 +326,7 @@ func buildFenceAgentParams(far *v1alpha1.FenceAgentsRemediation) ([]string, erro
 	fenceAgentParams = appendParamToSlice(fenceAgentParams, parameterActionName, parameterActionValue)
 
 	// append node parameters
-	nodeName := v1alpha1.NodeName(far.Name)
+	nodeName := v1alpha1.NodeName(getNodeName(far))
 	for paramName, nodeMap := range far.Spec.NodeParameters {
 		if nodeVal, isFound := nodeMap[nodeName]; isFound {
 			fenceAgentParams = appendParamToSlice(fenceAgentParams, paramName, nodeVal)

controllers/fenceagentsremediation_controller_test.go

@@ -163,12 +163,8 @@ var _ = Describe("FAR Controller", func() {
 		})
 
 		When("creating valid FAR CR", func() {
-			BeforeEach(func() {
-				node = utils.GetNode("", workerNode)
-				underTestFAR = getFenceAgentsRemediation(workerNode, fenceAgentIPMI, testShareParam, testNodeParam)
-			})
 
-			It("should have finalizer, taint, while the two VAs and one pod will be deleted", func() {
+			testSuccessfulRemediation := func() {
 				Eventually(func(g Gomega) {
 					g.Expect(storedCommand).To(ConsistOf([]string{
 						"fence_ipmilan",
@@ -180,7 +176,7 @@ var _ = Describe("FAR Controller", func() {
 						"--ipport=6233"}))
 				}, timeoutPreRemediation, pollInterval).Should(Succeed())
 
-				underTestFAR = verifyPreRemediationSucceed(underTestFAR, workerNode, defaultNamespace, &farRemediationTaint)
+				underTestFAR = verifyPreRemediationSucceed(underTestFAR, defaultNamespace, &farRemediationTaint)
 
 				By("Not having any test pod")
 				verifyPodDeleted(testPodName)
@@ -194,7 +190,23 @@ var _ = Describe("FAR Controller", func() {
 					conditionStatusPointer(metav1.ConditionTrue))  // SucceededTypeStatus
 				verifyEvent(corev1.EventTypeNormal, utils.EventReasonFenceAgentSucceeded, utils.EventMessageFenceAgentSucceeded)
 				verifyEvent(corev1.EventTypeNormal, utils.EventReasonNodeRemediationCompleted, utils.EventMessageNodeRemediationCompleted)
+			}
+			BeforeEach(func() {
+				node = utils.GetNode("", workerNode)
+				underTestFAR = getFenceAgentsRemediation(workerNode, fenceAgentIPMI, testShareParam, testNodeParam)
+			})
+			When("node name is stored in remediation name", func() {
+				It("should have finalizer, taint, while the two VAs and one pod will be deleted", testSuccessfulRemediation)
+			})
+			//remediation is created from escalation remediation supporting same kind template
+			When("node name is stored in remediation's annotation", func() {
+				BeforeEach(func() {
+					underTestFAR.Name = fmt.Sprintf("%s-%s", workerNode, "pseudo-random-test-sufix")
+					underTestFAR.Annotations = map[string]string{"remediation.medik8s.io/node-name": workerNode}
+				})
+				It("should have finalizer, taint, while the two VAs and one pod will be deleted", testSuccessfulRemediation)
 			})
+
 		})
 
 		When("creating invalid FAR CR Name", func() {
@@ -255,7 +267,7 @@ var _ = Describe("FAR Controller", func() {
 				})
 
 				It("should exit immediately without trying to update the status conditions", func() {
-					underTestFAR = verifyPreRemediationSucceed(underTestFAR, workerNode, defaultNamespace, &farRemediationTaint)
+					underTestFAR = verifyPreRemediationSucceed(underTestFAR, defaultNamespace, &farRemediationTaint)
 
 					By("Wait some retries")
 					Eventually(func() int {
@@ -285,7 +297,7 @@ var _ = Describe("FAR Controller", func() {
 				})
 
 				It("should exit immediately without trying to update the status conditions", func() {
-					underTestFAR = verifyPreRemediationSucceed(underTestFAR, workerNode, defaultNamespace, &farRemediationTaint)
+					underTestFAR = verifyPreRemediationSucceed(underTestFAR, defaultNamespace, &farRemediationTaint)
 
 					By("Deleting FAR CR")
 					Expect(k8sClient.Delete(context.Background(), underTestFAR)).To(Succeed())
@@ -308,7 +320,7 @@ var _ = Describe("FAR Controller", func() {
 				})
 
 				It("should retry the fence agent command as configured and update the status accordingly", func() {
-					underTestFAR = verifyPreRemediationSucceed(underTestFAR, workerNode, defaultNamespace, &farRemediationTaint)
+					underTestFAR = verifyPreRemediationSucceed(underTestFAR, defaultNamespace, &farRemediationTaint)
 
 					By("Still having one test pod")
 					verifyPodExists(testPodName)
@@ -338,7 +350,7 @@ var _ = Describe("FAR Controller", func() {
 				})
 
 				It("should stop Fence Agent execution and update the status accordingly", func() {
-					underTestFAR = verifyPreRemediationSucceed(underTestFAR, workerNode, defaultNamespace, &farRemediationTaint)
+					underTestFAR = verifyPreRemediationSucceed(underTestFAR, defaultNamespace, &farRemediationTaint)
 
 					By("Still having one test pod")
 					verifyPodExists(testPodName)
@@ -492,16 +504,16 @@ func verifyStatusCondition(far *v1alpha1.FenceAgentsRemediation, nodeName, condi
 }
 
 // verifyPreRemediationSucceed checks if the remediation CR already has a finazliaer and a remediation taint
-func verifyPreRemediationSucceed(underTestFAR *v1alpha1.FenceAgentsRemediation, nodeName, namespace string, taint *corev1.Taint) *v1alpha1.FenceAgentsRemediation {
+func verifyPreRemediationSucceed(underTestFAR *v1alpha1.FenceAgentsRemediation, namespace string, taint *corev1.Taint) *v1alpha1.FenceAgentsRemediation {
 	By("Searching for finalizer ")
-	Expect(k8sClient.Get(context.Background(), client.ObjectKey{Name: nodeName, Namespace: namespace}, underTestFAR)).To(Succeed())
+	Expect(k8sClient.Get(context.Background(), client.ObjectKey{Name: underTestFAR.GetName(), Namespace: namespace}, underTestFAR)).To(Succeed())
 	Expect(controllerutil.ContainsFinalizer(underTestFAR, v1alpha1.FARFinalizer)).To(BeTrue())
 	verifyEvent(corev1.EventTypeNormal, utils.EventReasonRemediationStarted, utils.EventMessageRemediationStarted)
 
 	By("Searching for remediation taint if we have a finalizer")
 	Eventually(func(g Gomega) {
 		node := &corev1.Node{}
-		g.Expect(k8sClient.Get(context.Background(), client.ObjectKey{Name: nodeName}, node)).To(Succeed())
+		g.Expect(k8sClient.Get(context.Background(), client.ObjectKey{Name: getNodeName(underTestFAR)}, node)).To(Succeed())
 		g.Expect(utils.TaintExists(node.Spec.Taints, taint)).To(BeTrue(), "remediation taint should exist")
 	}, timeoutPreRemediation, pollInterval).Should(Succeed())
 	verifyEvent(corev1.EventTypeNormal, utils.EventReasonAddRemediationTaint, utils.EventMessageAddRemediationTaint)

go.mod

@@ -4,7 +4,7 @@ go 1.20
 
 require (
 	github.com/go-logr/logr v1.4.1
-	github.com/medik8s/common v1.12.0
+	github.com/medik8s/common v1.17.0
 	github.com/onsi/ginkgo/v2 v2.14.0
 	github.com/onsi/gomega v1.30.0
 	github.com/openshift/api v0.0.0-20230621174358-ea40115b9fa6

go.sum

@@ -78,8 +78,8 @@ github.com/mailru/easyjson v0.7.7 h1:UGYAvKxe3sBsEDzO8ZeWOSlIQfWFlxbzLZe7hwFURr0
 github.com/mailru/easyjson v0.7.7/go.mod h1:xzfreul335JAWq5oZzymOObrkdz5UnU4kGfJJLY9Nlc=
 github.com/matttproud/golang_protobuf_extensions/v2 v2.0.0 h1:jWpvCLoY8Z/e3VKvlsiIGKtc+UG6U5vzxaoagmhXfyg=
 github.com/matttproud/golang_protobuf_extensions/v2 v2.0.0/go.mod h1:QUyp042oQthUoa9bqDv0ER0wrtXnBruoNd7aNjkbP+k=
-github.com/medik8s/common v1.12.0 h1:UJ5VS4rbo/K0snfuqRiYam1NhXTwo4Q7fTng34YSlmA=
-github.com/medik8s/common v1.12.0/go.mod h1:Q6YR2rgyiLl6ob4Mx2uDBmybAB3d94fImwoHPbeiE54=
+github.com/medik8s/common v1.17.0 h1:AmJKx0tzqGZF27Ot0A4ak85q0F0zqUkVyCvYmm67rtY=
+github.com/medik8s/common v1.17.0/go.mod h1:A9jYldC6PZcAuBowNNm712FqWdASB2ey5Vjp8MYN/PY=
 github.com/moby/spdystream v0.2.0 h1:cjW1zVyyoiM0T7b6UoySUFqzXMoqRckQtXwGPiBhOM8=
 github.com/moby/spdystream v0.2.0/go.mod h1:f7i0iNDQJ059oMTcWxx8MA/zKFIuD/lY+0GqbN2Wy8c=
 github.com/modern-go/concurrent v0.0.0-20180228061459-e0a39a4cb421/go.mod h1:6dJC0mAP4ikYIbvyc7fijjWJddQyLn8Ig3JB5CqoB9Q=