Merge pull request #95 from openshift-cherrypick-robot/cherry-pick-94…

…-to-release-4.10

BUG 2047642: [release-4.10] fix: use operator image for init containers

Dockerfile.combined

@@ -23,6 +23,10 @@ RUN CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -a -o metricsexporter cmd/met
 
 # vgmanager needs 'nsenter' and other basic linux utils to correctly function
 FROM centos:8
+
+# Install required utilities
+RUN dnf install -y openssl && dnf clean all
+
 WORKDIR /
 COPY --from=builder /workspace/manager .
 COPY --from=builder /workspace/vgmanager .

controllers/defaults.go

@@ -86,7 +86,6 @@ var (
 	CSIKubeletRootDir               = "/var/lib/kubelet/"
 	NodeContainerName               = "topolvm-node"
 	TopolvmNodeContainerHealthzName = "healthz"
-	auxImage                        = "registry.access.redhat.com/ubi8/ubi-minimal"
 	LvmdConfigFile                  = "/etc/topolvm/lvmd.yaml"
 
 	// topoLVM Node resource requests/limits

controllers/lvmcluster_controller.go

@@ -19,9 +19,12 @@ package controllers
 import (
 	"context"
 	"fmt"
+	"os"
 
+	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/types"
 
 	"github.com/go-logr/logr"
 	secv1client "github.com/openshift/client-go/security/clientset/versioned/typed/security/v1"
@@ -57,6 +60,7 @@ type LVMClusterReconciler struct {
 	ClusterType    ClusterType
 	SecurityClient secv1client.SecurityV1Interface
 	Namespace      string
+	ImageName      string
 }
 
 //+kubebuilder:rbac:groups=lvm.topolvm.io,resources=lvmclusters,verbs=get;list;watch;create;update;patch;delete
@@ -101,6 +105,13 @@ func (r *LVMClusterReconciler) Reconcile(ctx context.Context, req ctrl.Request)
 		r.Log.Error(err, "failed to check cluster type")
 		return ctrl.Result{}, err
 	}
+
+	err = r.getRunningPodImage(ctx)
+	if err != nil {
+		r.Log.Error(err, "failed to get operator image")
+		return ctrl.Result{}, err
+	}
+
 	result, reconcileError := r.reconcile(ctx, lvmCluster)
 
 	statusError := r.updateLVMClusterStatus(ctx, lvmCluster)
@@ -290,3 +301,33 @@ func (r *LVMClusterReconciler) checkIfOpenshift(ctx context.Context) error {
 func IsOpenshift(r *LVMClusterReconciler) bool {
 	return r.ClusterType == ClusterTypeOCP
 }
+
+// getRunningPodImage gets the operator image and set it in reconciler struct
+func (r *LVMClusterReconciler) getRunningPodImage(ctx context.Context) error {
+
+	if r.ImageName == "" {
+		// 'POD_NAME' and 'POD_NAMESPACE' are set in env of lvm-operator when running as a container
+		podName := os.Getenv("POD_NAME")
+		if podName == "" {
+			return fmt.Errorf("failed to get pod name env variable")
+		}
+
+		pod := &corev1.Pod{}
+		if err := r.Get(ctx, types.NamespacedName{Name: podName, Namespace: r.Namespace}, pod); err != nil {
+			return fmt.Errorf("failed to get pod %s in namespace %s", podName, r.Namespace)
+		}
+
+		for _, c := range pod.Spec.Containers {
+			if c.Name == LVMOperatorContainerName {
+				r.ImageName = c.Image
+				return nil
+			}
+		}
+
+		return fmt.Errorf("failed to get container image for %s in pod %s", LVMOperatorContainerName, podName)
+
+	}
+
+	return nil
+
+}

controllers/suite_test.go

@@ -61,6 +61,7 @@ const (
 	testLvmClusterName      = "test-lvmcluster"
 	testLvmClusterNamespace = "openshift-storage"
 	testDeviceClassName     = "test"
+	testImageName           = "test"
 )
 
 var _ = BeforeSuite(func() {
@@ -103,6 +104,7 @@ var _ = BeforeSuite(func() {
 		SecurityClient: secv1client.NewForConfigOrDie(k8sManager.GetConfig()),
 		Namespace:      testLvmClusterNamespace,
 		Log:            ctrl.Log.WithName("controllers").WithName("LvmCluster"),
+		ImageName:      testImageName,
 	}).SetupWithManager(k8sManager)
 	Expect(err).ToNot(HaveOccurred())
 

controllers/topolvm_controller.go

@@ -34,7 +34,7 @@ func (c topolvmController) getName() string {
 func (c topolvmController) ensureCreated(r *LVMClusterReconciler, ctx context.Context, lvmCluster *lvmv1alpha1.LVMCluster) error {
 
 	// get the desired state of topolvm controller deployment
-	desiredDeployment := getControllerDeployment(lvmCluster, r.Namespace)
+	desiredDeployment := getControllerDeployment(lvmCluster, r.Namespace, r.ImageName)
 	existingDeployment := &appsv1.Deployment{
 		ObjectMeta: metav1.ObjectMeta{
 			Name:      desiredDeployment.Name,
@@ -105,8 +105,7 @@ func (c topolvmController) setTopolvmControllerDesiredState(existing, desired *a
 	return nil
 }
 
-func getControllerDeployment(lvmCluster *lvmv1alpha1.LVMCluster, namespace string) *appsv1.Deployment {
-
+func getControllerDeployment(lvmCluster *lvmv1alpha1.LVMCluster, namespace string, initImage string) *appsv1.Deployment {
 	// Topolvm CSI Controller Deployment
 	var replicas int32 = 1
 	volumes := []corev1.Volume{
@@ -116,7 +115,7 @@ func getControllerDeployment(lvmCluster *lvmv1alpha1.LVMCluster, namespace strin
 
 	// TODO: Remove custom generation of TLS certs, current it's being used in topolvm controller manager
 	initContainers := []corev1.Container{
-		*getInitContainer(),
+		*getInitContainer(initImage),
 	}
 
 	// get all containers that are part of csi controller deployment
@@ -158,11 +157,11 @@ func getControllerDeployment(lvmCluster *lvmv1alpha1.LVMCluster, namespace strin
 	return controllerDeployment
 }
 
-func getInitContainer() *corev1.Container {
+func getInitContainer(initImage string) *corev1.Container {
 
 	// generation of tls certs
 	command := []string{
-		"sh",
+		"/usr/bin/bash",
 		"-c",
 		"openssl req -nodes -x509 -newkey rsa:4096 -subj '/DC=self_signed_certificate' -keyout /certs/tls.key -out /certs/tls.crt -days 3650",
 	}
@@ -173,7 +172,7 @@ func getInitContainer() *corev1.Container {
 
 	ssCertGenerator := &corev1.Container{
 		Name:         "self-signed-cert-generator",
-		Image:        "alpine/openssl",
+		Image:        initImage,
 		Command:      command,
 		VolumeMounts: volumeMounts,
 	}

controllers/topolvm_node.go

@@ -49,7 +49,8 @@ func (n topolvmNode) ensureCreated(r *LVMClusterReconciler, ctx context.Context,
 	unitLogger := r.Log.WithValues("topolvmNode", n.getName())
 
 	// get desired daemonSet spec
-	dsTemplate := getNodeDaemonSet(lvmCluster, r.Namespace)
+	dsTemplate := getNodeDaemonSet(lvmCluster, r.Namespace, r.ImageName)
+
 	// create desired daemonSet or update mutable fields on existing one
 	ds := &appsv1.DaemonSet{
 		ObjectMeta: metav1.ObjectMeta{
@@ -127,7 +128,8 @@ func (n topolvmNode) updateStatus(r *LVMClusterReconciler, ctx context.Context,
 	return nil
 }
 
-func getNodeDaemonSet(lvmCluster *lvmv1alpha1.LVMCluster, namespace string) *appsv1.DaemonSet {
+func getNodeDaemonSet(lvmCluster *lvmv1alpha1.LVMCluster, namespace string, initImage string) *appsv1.DaemonSet {
+
 	hostPathDirectory := corev1.HostPathDirectory
 	hostPathDirectoryOrCreateType := corev1.HostPathDirectoryOrCreate
 	storageMedium := corev1.StorageMediumMemory
@@ -163,7 +165,7 @@ func getNodeDaemonSet(lvmCluster *lvmv1alpha1.LVMCluster, namespace string) *app
 				EmptyDir: &corev1.EmptyDirVolumeSource{Medium: storageMedium}}},
 	}
 
-	initContainers := []corev1.Container{*getNodeInitContainer()}
+	initContainers := []corev1.Container{*getNodeInitContainer(initImage)}
 	containers := []corev1.Container{*getLvmdContainer(), *getNodeContainer(), *getCsiRegistrarContainer(), *getNodeLivenessProbeContainer()}
 
 	// Affinity and tolerations
@@ -210,9 +212,9 @@ func getNodeDaemonSet(lvmCluster *lvmv1alpha1.LVMCluster, namespace string) *app
 	return nodeDaemonSet
 }
 
-func getNodeInitContainer() *corev1.Container {
+func getNodeInitContainer(initImage string) *corev1.Container {
 	command := []string{
-		"sh",
+		"/usr/bin/bash",
 		"-c",
 		fmt.Sprintf("until [ -f %s ]; do echo waiting for lvmd config file; sleep 5; done", LvmdConfigFile),
 	}
@@ -223,7 +225,7 @@ func getNodeInitContainer() *corev1.Container {
 
 	fileChecker := &corev1.Container{
 		Name:         "file-checker",
-		Image:        auxImage,
+		Image:        initImage,
 		Command:      command,
 		VolumeMounts: volumeMounts,
 	}

controllers/vgmanager.go

@@ -43,13 +43,10 @@ func (v vgManager) ensureCreated(r *LVMClusterReconciler, ctx context.Context, l
 	unitLogger := r.Log.WithValues("resourceManager", v.getName())
 
 	// get desired daemonset spec
-	dsTemplate, err := newVGManagerDaemonset(r, ctx, lvmCluster)
-	if err != nil {
-		return fmt.Errorf("failed to get new VGManager Daemonset due to %v", err)
-	}
+	dsTemplate := newVGManagerDaemonset(lvmCluster, r.Namespace, r.ImageName)
 
 	// controller reference
-	err = ctrl.SetControllerReference(lvmCluster, &dsTemplate, r.Scheme)
+	err := ctrl.SetControllerReference(lvmCluster, &dsTemplate, r.Scheme)
 	if err != nil {
 		return fmt.Errorf("failed to set controller reference on vgManager daemonset %q. %v", dsTemplate.Name, err)
 	}

controllers/vgmanager_daemonset.go

@@ -17,15 +17,12 @@ limitations under the License.
 package controllers
 
 import (
-	"context"
-	"fmt"
 	"os"
 
 	lvmv1alpha1 "github.com/red-hat-storage/lvm-operator/api/v1alpha1"
 	appsv1 "k8s.io/api/apps/v1"
 	corev1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-	"k8s.io/apimachinery/pkg/types"
 )
 
 var (
@@ -117,7 +114,7 @@ var (
 )
 
 // newVGManagerDaemonset returns the desired vgmanager daemonset for a given LVMCluster
-func newVGManagerDaemonset(r *LVMClusterReconciler, ctx context.Context, lvmCluster *lvmv1alpha1.LVMCluster) (appsv1.DaemonSet, error) {
+func newVGManagerDaemonset(lvmCluster *lvmv1alpha1.LVMCluster, namespace string, vgImage string) appsv1.DaemonSet {
 	// aggregate nodeSelector and tolerations from all deviceClasses
 	nodeSelector, tolerations := extractNodeSelectorAndTolerations(lvmCluster)
 	volumes := []corev1.Volume{LVMDConfVol, DevHostDirVol, UDevHostDirVol, SysHostDirVol}
@@ -128,17 +125,10 @@ func newVGManagerDaemonset(r *LVMClusterReconciler, ctx context.Context, lvmClus
 	// try to get vgmanager image from env and on absence get from running pod
 	// TODO: investigate why storing this env in a variable is failing tests
 	image := os.Getenv("VGMANAGER_IMAGE")
-	var err error
 	if image == "" {
-		image, err = getRunningPodImage(r, ctx)
-		if err != nil {
-			r.Log.Error(err, "failed to get image from running operator")
-			return appsv1.DaemonSet{}, err
-		}
+		image = vgImage
 	}
 
-	r.Log.Info("creating VG manager deployment", "image", image)
-
 	command := []string{
 		"/vgmanager",
 	}
@@ -187,7 +177,7 @@ func newVGManagerDaemonset(r *LVMClusterReconciler, ctx context.Context, lvmClus
 	ds := appsv1.DaemonSet{
 		ObjectMeta: metav1.ObjectMeta{
 			Name:      VGManagerUnit,
-			Namespace: r.Namespace,
+			Namespace: namespace,
 			Labels:    labels,
 		},
 		Spec: appsv1.DaemonSetSpec{
@@ -211,28 +201,5 @@ func newVGManagerDaemonset(r *LVMClusterReconciler, ctx context.Context, lvmClus
 
 	// set nodeSelector
 	setDaemonsetNodeSelector(nodeSelector, &ds)
-	return ds, nil
-}
-
-func getRunningPodImage(r *LVMClusterReconciler, ctx context.Context) (string, error) {
-
-	// 'POD_NAME' and 'POD_NAMESPACE' are set in env of lvm-operator when running as a container
-	podName := os.Getenv("POD_NAME")
-	if podName == "" {
-		return "", fmt.Errorf("failed to get pod name env variable")
-	}
-
-	pod := &corev1.Pod{}
-	if err := r.Get(ctx, types.NamespacedName{Name: podName, Namespace: r.Namespace}, pod); err != nil {
-		return "", fmt.Errorf("failed to get pod %s in namespace %s", podName, r.Namespace)
-	}
-
-	for _, c := range pod.Spec.Containers {
-		if c.Name == LVMOperatorContainerName {
-			return c.Image, nil
-		}
-	}
-
-	return "", fmt.Errorf("failed to get container image for %s in pod %s", LVMOperatorContainerName, podName)
-
+	return ds
 }