core: add ROOK_REVISION_HISTORY_LIMIT operator setting

This adds an operator config setting ROOK_REVISION_HISTORY_LIMIT
defaulting to kubernetes'value for RevisionHistoryLimit.

If configured, the provided value will be used as RevisionHistoryLimit

for all Deployments rook creates.

Fixes: #12722

Signed-off-by: Michael Adam <obnox@samba.org>
(cherry picked from commit ab8fd90aa6)

deploy/examples/operator-openshift.yaml

@@ -638,6 +638,8 @@ data:
 
   # (Optional) QPS to use while communicating with the kubernetes apiserver.
   # CSI_KUBE_API_QPS: "5.0"
+  # RevisionHistoryLimit value for all deployments created by rook.
+  # ROOK_REVISION_HISTORY_LIMIT: "3"
 ---
 # The deployment for the rook operator
 # OLM: BEGIN OPERATOR DEPLOYMENT

deploy/examples/operator.yaml

@@ -568,6 +568,8 @@ data:
 
   # (Optional) QPS to use while communicating with the kubernetes apiserver.
   # CSI_KUBE_API_QPS: "5.0"
+  # RevisionHistoryLimit value for all deployments created by rook.
+  # ROOK_REVISION_HISTORY_LIMIT: "3"
 ---
 # OLM: BEGIN OPERATOR DEPLOYMENT
 apiVersion: apps/v1

pkg/operator/ceph/cluster/mgr/spec.go

@@ -128,6 +128,7 @@ func (c *Cluster) makeDeployment(mgrConfig *mgrConfig) (*apps.Deployment, error)
 			Labels:    c.getPodLabels(mgrConfig, true),
 		},
 		Spec: apps.DeploymentSpec{
+			RevisionHistoryLimit: controller.RevisionHistoryLimit(),
 			Selector: &metav1.LabelSelector{
 				MatchLabels: c.getPodLabels(mgrConfig, false),
 			},

pkg/operator/ceph/cluster/mon/spec.go

@@ -108,6 +108,7 @@ func (c *Cluster) makeDeployment(monConfig *monConfig, canary bool) (*apps.Deplo
 	}
 	replicaCount := int32(1)
 	d.Spec = apps.DeploymentSpec{
+		RevisionHistoryLimit: controller.RevisionHistoryLimit(),
 		Selector: &metav1.LabelSelector{
 			MatchLabels: c.getLabels(monConfig, canary, false),
 		},

pkg/operator/ceph/cluster/nodedaemon/exporter.go

@@ -77,6 +77,7 @@ func (r *ReconcileNode) createOrUpdateCephExporter(node corev1.Node, tolerations
 			Namespace: cephCluster.GetNamespace(),
 		},
 	}
+	deploy.Spec.RevisionHistoryLimit = controller.RevisionHistoryLimit()
 	err := controllerutil.SetControllerReference(&cephCluster, deploy, r.scheme)
 	if err != nil {
 		return controllerutil.OperationResultNone, errors.Errorf("failed to set owner reference of ceph-exporter deployment %q", deploy.Name)

pkg/operator/ceph/cluster/osd/spec.go

@@ -713,6 +713,7 @@ func (c *Cluster) makeDeployment(osdProps osdProperties, osd *OSDInfo, provision
 					OsdIdLabelKey:       fmt.Sprintf("%d", osd.ID),
 				},
 			},
+			RevisionHistoryLimit: controller.RevisionHistoryLimit(),
 			Strategy: apps.DeploymentStrategy{
 				Type: apps.RecreateDeploymentStrategyType,
 			},

pkg/operator/ceph/cluster/rbd/spec.go

@@ -82,6 +82,7 @@ func (r *ReconcileCephRBDMirror) makeDeployment(daemonConfig *daemonConfig, rbdM
 			Labels:      controller.CephDaemonAppLabels(AppName, rbdMirror.Namespace, config.RbdMirrorType, daemonConfig.DaemonID, rbdMirror.Name, "cephrbdmirrors.ceph.rook.io", true),
 		},
 		Spec: apps.DeploymentSpec{
+			RevisionHistoryLimit: controller.RevisionHistoryLimit(),
 			Selector: &metav1.LabelSelector{
 				MatchLabels: podSpec.Labels,
 			},

pkg/operator/ceph/controller.go

@@ -133,6 +133,7 @@ func (r *ReconcileConfig) reconcile(request reconcile.Request) (reconcile.Result
 
 	opcontroller.SetAllowLoopDevices(r.config.Parameters)
 	opcontroller.SetEnforceHostNetwork(r.config.Parameters)
+	opcontroller.SetRevisionHistoryLimit(r.config.Parameters)
 
 	logger.Infof("%s done reconciling", controllerName)
 	return reconcile.Result{}, nil

pkg/operator/ceph/controller/controller_utils.go

@@ -54,6 +54,9 @@ const (
 	enforceHostNetworkSettingName  string = "ROOK_ENFORCE_HOST_NETWORK"
 	enforceHostNetworkDefaultValue string = "false"
 
+	revisionHistoryLimitSettingName  string = "ROOK_REVISION_HISTORY_LIMIT"
+	revisionHistoryLimitDefaultValue string = ""
+
 	// UninitializedCephConfigError refers to the error message printed by the Ceph CLI when there is no ceph configuration file
 	// This typically is raised when the operator has not finished initializing
 	UninitializedCephConfigError = "error calling conf_read_file"
@@ -86,7 +89,8 @@ var (
 	OperatorCephBaseImageVersion string
 
 	// loopDevicesAllowed indicates whether loop devices are allowed to be used
-	loopDevicesAllowed = false
+	loopDevicesAllowed          = false
+	revisionHistoryLimit *int32 = nil
 )
 
 func DiscoveryDaemonEnabled(data map[string]string) bool {
@@ -133,6 +137,26 @@ func EnforceHostNetwork() bool {
 	return cephv1.EnforceHostNetwork()
 }
 
+func SetRevisionHistoryLimit(data map[string]string) {
+	strval := k8sutil.GetValue(data, revisionHistoryLimitSettingName, revisionHistoryLimitDefaultValue)
+	if strval != "" {
+		numval, err := strconv.ParseInt(strval, 10, 32)
+		if err != nil {
+			logger.Warningf("failed to parse value %q for %q. assuming default value.", strval, revisionHistoryLimitSettingName)
+			revisionHistoryLimit = nil
+			return
+
+		}
+		limit := int32(numval)
+		revisionHistoryLimit = &limit
+	}
+
+}
+
+func RevisionHistoryLimit() *int32 {
+	return revisionHistoryLimit
+}
+
 // canIgnoreHealthErrStatusInReconcile determines whether a status of HEALTH_ERR in the CephCluster can be ignored safely.
 func canIgnoreHealthErrStatusInReconcile(cephCluster cephv1.CephCluster, controllerName string) bool {
 	// Get a list of all the keys causing the HEALTH_ERR status.

pkg/operator/ceph/controller/controller_utils_test.go

@@ -133,6 +133,33 @@ func TestSetEnforceHostNetwork(t *testing.T) {
 	assert.False(t, EnforceHostNetwork())
 }
 
+func TestSetRevisionHistoryLimit(t *testing.T) {
+	opConfig := map[string]string{}
+	t.Run("ROOK_REVISION_HISTORY_LIMIT: test default value", func(t *testing.T) {
+		SetRevisionHistoryLimit(opConfig)
+		assert.Nil(t, RevisionHistoryLimit())
+	})
+
+	var value string = "foo"
+	t.Run("ROOK_REVISION_HISTORY_LIMIT: test invalid value 'foo'", func(t *testing.T) {
+		opConfig[revisionHistoryLimitSettingName] = value
+		SetRevisionHistoryLimit(opConfig)
+		assert.Nil(t, RevisionHistoryLimit())
+	})
+
+	t.Run("ROOK_REVISION_HISTORY_LIMIT: test empty string value", func(t *testing.T) {
+		value = ""
+		opConfig[revisionHistoryLimitSettingName] = value
+		SetRevisionHistoryLimit(opConfig)
+		assert.Nil(t, RevisionHistoryLimit())
+	})
+	t.Run("ROOK_REVISION_HISTORY_LIMIT:  test valig value '10'", func(t *testing.T) {
+		value = "10"
+		opConfig[revisionHistoryLimitSettingName] = value
+		SetRevisionHistoryLimit(opConfig)
+		assert.Equal(t, int32(10), *RevisionHistoryLimit())
+	})
+}
 func TestIsReadyToReconcile(t *testing.T) {
 	scheme := scheme.Scheme
 	scheme.AddKnownTypes(cephv1.SchemeGroupVersion, &cephv1.CephCluster{}, &cephv1.CephClusterList{})

pkg/operator/ceph/file/mds/spec.go

@@ -94,8 +94,9 @@ func (c *Cluster) makeDeployment(mdsConfig *mdsConfig, fsNamespacedname types.Na
 			Selector: &metav1.LabelSelector{
 				MatchLabels: c.podLabels(mdsConfig, false),
 			},
-			Template: podSpec,
-			Replicas: &replicas,
+			RevisionHistoryLimit: controller.RevisionHistoryLimit(),
+			Template:             podSpec,
+			Replicas:             &replicas,
 			Strategy: apps.DeploymentStrategy{
 				Type: apps.RecreateDeploymentStrategyType,
 			},

pkg/operator/ceph/file/mirror/spec.go

@@ -79,6 +79,7 @@ func (r *ReconcileFilesystemMirror) makeDeployment(daemonConfig *daemonConfig, f
 			Annotations: fsMirror.Spec.Annotations,
 			Labels:      controller.CephDaemonAppLabels(AppName, fsMirror.Namespace, config.FilesystemMirrorType, userID, fsMirror.Name, "cephfilesystemmirrors.ceph.rook.io", true)},
 		Spec: apps.DeploymentSpec{
+			RevisionHistoryLimit: controller.RevisionHistoryLimit(),
 			Selector: &metav1.LabelSelector{
 				MatchLabels: podSpec.Labels,
 			},

pkg/operator/ceph/nfs/spec.go

@@ -191,6 +191,7 @@ func (r *ReconcileCephNFS) makeDeployment(nfs *cephv1.CephNFS, cfg daemonConfig)
 	// Multiple replicas of the nfs service would be handled by creating a service and a new deployment for each one, rather than increasing the pod count here
 	replicas := int32(1)
 	deployment.Spec = apps.DeploymentSpec{
+		RevisionHistoryLimit: controller.RevisionHistoryLimit(),
 		Selector: &metav1.LabelSelector{
 			MatchLabels: getLabels(nfs, cfg.ID, false),
 		},

pkg/operator/ceph/object/cosi/spec.go

@@ -43,7 +43,6 @@ func createCephCOSIDriverDeployment(cephCOSIDriver *cephv1.CephCOSIDriver) (*app
 	replica := int32(1)
 	minReadySeconds := int32(30)
 	progressDeadlineSeconds := int32(600)
-	revisionHistoryLimit := int32(3)
 
 	cephcosidriverDeployment := &appsv1.Deployment{
 		ObjectMeta: metav1.ObjectMeta{
@@ -52,7 +51,8 @@ func createCephCOSIDriverDeployment(cephCOSIDriver *cephv1.CephCOSIDriver) (*app
 			Labels:    getCOSILabels(cephCOSIDriver.Name, cephCOSIDriver.Namespace),
 		},
 		Spec: appsv1.DeploymentSpec{
-			Replicas: &replica,
+			RevisionHistoryLimit: controller.RevisionHistoryLimit(),
+			Replicas:             &replica,
 			Selector: &metav1.LabelSelector{
 				MatchLabels: getCOSILabels(cephCOSIDriver.Name, cephCOSIDriver.Namespace),
 			},
@@ -60,7 +60,6 @@ func createCephCOSIDriverDeployment(cephCOSIDriver *cephv1.CephCOSIDriver) (*app
 			Strategy:                strategy,
 			MinReadySeconds:         minReadySeconds,
 			ProgressDeadlineSeconds: &progressDeadlineSeconds,
-			RevisionHistoryLimit:    &revisionHistoryLimit,
 		},
 	}
 

pkg/operator/ceph/object/spec.go

@@ -117,6 +117,7 @@ func (c *clusterConfig) createDeployment(rgwConfig *rgwConfig) (*apps.Deployment
 			Labels:    getLabels(c.store.Name, c.store.Namespace, true),
 		},
 		Spec: apps.DeploymentSpec{
+			RevisionHistoryLimit: controller.RevisionHistoryLimit(),
 			Selector: &metav1.LabelSelector{
 				MatchLabels: getLabels(c.store.Name, c.store.Namespace, false),
 			},

pkg/operator/discover/discover.go

@@ -102,6 +102,7 @@ func (d *Discover) createDiscoverDaemonSet(ctx context.Context, namespace, disco
 			Labels: getLabels(),
 		},
 		Spec: apps.DaemonSetSpec{
+			RevisionHistoryLimit: opcontroller.RevisionHistoryLimit(),
 			Selector: &metav1.LabelSelector{
 				MatchLabels: map[string]string{
 					"app": discoverDaemonsetName,