Merge pull request #198 from cscetbon/metrics

Add custom metrics to operator

MAINTAINERS

@@ -4,5 +4,7 @@
 #     Individual's name <submission email address> (@GITHUB_HANDLE)
 
 Sébastien Allamand <sebastien.allamand@orange.com> (@allamand)
-Cyril Scetbon <cscetbon.ext@orange.com> (@scetbon)
-Jean Armel Luce <jeanarmel.luce@orange.com> (@jal06)
+Cyril Scetbon <cscetbon.ext@orange.com> (@cscetbon)
+Jean Armel Luce <jeanarmel.luce@orange.com> (@jal06)
+Alexandre Guitton <aguitton.ext@orange.com> (@erdrix)
+Franck Dehay <franck.dehay@orange.com> (@fdehay)

cmd/manager/main.go

@@ -16,6 +16,7 @@ package main
 
 import (
 	"context"
+	"fmt"
 	"os"
 	"runtime"
 	"strconv"
@@ -29,6 +30,7 @@ import (
 	"github.com/Orange-OpenSource/casskop/pkg/apis"
 	api "github.com/Orange-OpenSource/casskop/pkg/apis/db/v1alpha1"
 	"github.com/Orange-OpenSource/casskop/pkg/controller"
+	"github.com/Orange-OpenSource/casskop/pkg/controller/cassandracluster"
 	"github.com/Orange-OpenSource/casskop/version"
 	"github.com/operator-framework/operator-sdk/pkg/k8sutil"
 	"github.com/operator-framework/operator-sdk/pkg/leader"
@@ -37,6 +39,7 @@ import (
 	sdkVersion "github.com/operator-framework/operator-sdk/version"
 	"github.com/sirupsen/logrus"
 	v1 "k8s.io/api/core/v1"
+	prometheusMetrics "sigs.k8s.io/controller-runtime/pkg/metrics"
 
 	"sigs.k8s.io/controller-runtime/pkg/client/config"
 	"sigs.k8s.io/controller-runtime/pkg/manager"
@@ -101,37 +104,12 @@ func getResyncPeriod() int {
 }
 
 func main() {
-	/*
-		// Add the zap logger flag set to the CLI. The flag set must
-		// be added before calling pflag.Parse().
-		pflag.CommandLine.AddFlagSet(zap.FlagSet())
-
-		// avoid glog to write to /tmp
-		flag.Set("logtostderr", "true")
-
-		// Add flags registered by imported packages (e.g. glog and
-		// controller-runtime)
-		pflag.CommandLine.AddGoFlagSet(flag.CommandLine)
-
-		pflag.Parse()
-	*/
-	// Use a zap logr.Logger implementation. If none of the zap
-	// flags are configured (or if the zap flag set is not being
-	// used), this defaults to a production zap logger.
-	//
-	// The logger instantiated here can be changed to any logger
-	// implementing the logr.Logger interface. This logger will
-	// be propagated through the whole operator, generating
-	// uniform and structured logs.
-	//logf.SetLogger(logf.ZapLogger(false))
-
 	logType, found := os.LookupEnv("LOG_TYPE")
 	if found && logType == "json" {
 		logrus.SetFormatter(&logrus.JSONFormatter{})
 	}
-	//logrus.SetOutput(os.Stdout)
+
 	logrus.SetLevel(getLogLevel())
-	//	resyncPeriod := getResyncPeriod()
 
 	printVersion()
 
@@ -171,8 +149,8 @@ func main() {
 
 	// Create a new Cmd to provide shared dependencies and start components
 	mgr, err := manager.New(cfg, manager.Options{
-		Namespace: namespace,
-		//MetricsBindAddress: fmt.Sprintf("%s:%d", metricsHost, metricsPort),
+		Namespace:          namespace,
+		MetricsBindAddress: fmt.Sprintf("%s:%d", metricsHost, metricsPort),
 	})
 	if err != nil {
 		logrus.Error(err)
@@ -181,6 +159,8 @@ func main() {
 
 	logrus.Info("Registering Components.")
 
+	prometheusMetrics.Registry.MustRegister(cassandracluster.ClusterActionMetric, cassandracluster.ClusterPhaseMetric)
+
 	// Setup Scheme for all resources
 	if err := apis.AddToScheme(mgr.GetScheme()); err != nil {
 		logrus.Error(err)
@@ -195,10 +175,11 @@ func main() {
 
 	// Create Service object to expose the metrics port.
 	servicePorts := []v1.ServicePort{
-		{Name: "metricsPort", TargetPort: intstr.FromInt(int(metricsPort))},
+		{Name: metrics.OperatorPortName, Port: metricsPort,
+			TargetPort: intstr.IntOrString{Type: intstr.Int, IntVal: metricsPort}},
 	}
-	_, err = metrics.CreateMetricsService(ctx, cfg, servicePorts)
-	if err != nil {
+
+	if _, err := metrics.CreateMetricsService(ctx, cfg, servicePorts); err != nil {
 		logrus.Info(err.Error())
 	}
 

go.mod

@@ -34,6 +34,7 @@ require (
 	github.com/maxbrunsfeld/counterfeiter v0.0.0-20181017030959-1aadac120687 // indirect
 	github.com/operator-framework/operator-sdk v0.15.0
 	github.com/pelletier/go-toml v1.3.0 // indirect
+	github.com/prometheus/client_golang v1.2.1
 	github.com/r3labs/diff v0.0.0-20190801153147-a71de73c46ad
 	github.com/shurcooL/sanitized_anchor_name v1.0.0 // indirect
 	github.com/sirupsen/logrus v1.4.2

go.sum

@@ -1666,6 +1666,7 @@ k8s.io/kube-openapi v0.0.0-20191107075043-30be4d16710a/go.mod h1:1TqjTSzOxsLGIKf
 k8s.io/kube-proxy v0.0.0-20191016114407-2e83b6f20229/go.mod h1:2Hxci1uzXO5ipP0h9n2+h18fvNkBTpYlckk5dOPu8zg=
 k8s.io/kube-scheduler v0.0.0-20191016114748-65049c67a58b/go.mod h1:BgDUHHC5Wl0xcBUQgo2XEprE5nG5i9tlRR4iNgEFbL0=
 k8s.io/kube-state-metrics v1.6.0/go.mod h1:84+q9aGVQPzXYGgtvyhZr/fSI6BdLsbPWXn37RASc9k=
+k8s.io/kube-state-metrics v1.7.2 h1:6vdtgXrrRRMSgnyDmgua+qvgCYv954JNfxXAtDkeLVQ=
 k8s.io/kube-state-metrics v1.7.2/go.mod h1:U2Y6DRi07sS85rmVPmBFlmv+2peBcL8IWGjM+IjYA/E=
 k8s.io/kubectl v0.0.0-20191016120415-2ed914427d51/go.mod h1:gL826ZTIfD4vXTGlmzgTbliCAT9NGiqpCqK2aNYv5MQ=
 k8s.io/kubelet v0.0.0-20191016114556-7841ed97f1b2/go.mod h1:SBvrtLbuePbJygVXGGCMtWKH07+qrN2dE1iMnteSG8E=

helm/cassandra-operator/templates/service.yaml

@@ -14,6 +14,6 @@ spec:
     app: {{ template "cassandra-operator.name" . }}
   ports:
   - name: metrics
-    port: 9710
+    port: 8383
     protocol: TCP
 {{- end }}

multi-casskop/pkg/controller/multi-casskop/cmc_utils.go

@@ -2,6 +2,7 @@ package multicasskop
 
 import (
 	"context"
+
 	"github.com/Orange-OpenSource/casskop/multi-casskop/pkg/controller/multi-casskop/models"
 
 	ccv1 "github.com/Orange-OpenSource/casskop/pkg/apis/db/v1alpha1"
@@ -14,7 +15,7 @@ import (
 // ReadyCassandraCluster
 // return true if CassandraCluster it Done and Running
 func (r *reconciler) ReadyCassandraCluster(cc *ccv1.CassandraCluster) bool {
-	if cc.Status.Phase != ccv1.ClusterPhaseRunning || cc.Status.LastClusterActionStatus != ccv1.StatusDone {
+	if cc.Status.Phase != ccv1.ClusterPhaseRunning.Name || cc.Status.LastClusterActionStatus != ccv1.StatusDone {
 		return false
 	}
 	return true

pkg/apis/db/v1alpha1/cassandracluster_types.go

@@ -52,12 +52,36 @@ const (
 	DefaultUserID int64 = 999
 )
 
+type ClusterStateInfo struct {
+	Id   float64
+	Name string
+}
+
+var (
+	//Cluster phases
+	ClusterPhaseInitial = ClusterStateInfo{1, "Initializing"}
+	ClusterPhaseRunning = ClusterStateInfo{2, "Running"}
+	ClusterPhasePending = ClusterStateInfo{3, "Pending"}
+
+	//Available actions
+	ActionUpdateConfigMap   = ClusterStateInfo{1, "UpdateConfigMap"}
+	ActionUpdateDockerImage = ClusterStateInfo{2, "UpdateDockerImage"}
+	ActionUpdateSeedList    = ClusterStateInfo{3, "UpdateSeedList"}
+	ActionRollingRestart    = ClusterStateInfo{4, "RollingRestart"}
+	ActionUpdateResources   = ClusterStateInfo{5, "UpdateResources"}
+	ActionUpdateStatefulSet = ClusterStateInfo{6, "UpdateStatefulSet"}
+	ActionScaleUp           = ClusterStateInfo{7, "ScaleUp"}
+	ActionScaleDown         = ClusterStateInfo{8, "ScaleDown"}
+
+	ActionDeleteDC   = ClusterStateInfo{9, "ActionDeleteDC"}
+	ActionDeleteRack = ClusterStateInfo{10, "ActionDeleteRack"}
+
+	ActionCorrectCRDConfig = ClusterStateInfo{11, "CorrectCRDConfig"} //The Operator has correct a bad CRD configuration
+
+)
+
 const (
 	AnnotationLastApplied string = "cassandraclusters.db.orange.com/last-applied-configuration"
-	//Phase du Cluster
-	ClusterPhaseInitial string = "Initializing"
-	ClusterPhaseRunning string = "Running"
-	ClusterPhasePending string = "Pending"
 
 	StatusOngoing     string = "Ongoing"    // The Action is Ongoing
 	StatusDone        string = "Done"       // The Action id Done
@@ -68,21 +92,6 @@ const (
 	StatusManual      string = "Manual"
 	StatusError       string = "Error"
 
-	//Available actions
-	ActionUpdateConfigMap   string = "UpdateConfigMap"
-	ActionUpdateDockerImage string = "UpdateDockerImage"
-	ActionUpdateSeedList    string = "UpdateSeedList"
-	ActionRollingRestart    string = "RollingRestart"
-	ActionUpdateResources   string = "UpdateResources"
-	ActionUpdateStatefulSet string = "UpdateStatefulSet"
-	ActionScaleUp           string = "ScaleUp"
-	ActionScaleDown         string = "ScaleDown"
-
-	ActionDeleteDC   string = "ActionDeleteDC"
-	ActionDeleteRack string = "ActionDeleteRack"
-
-	ActionCorrectCRDConfig string = "CorrectCRDConfig" //The Operator has correct a bad CRD configuration
-
 	//List of Pods Operations
 	OperationUpgradeSSTables string = "upgradesstables"
 	OperationCleanup         string = "cleanup"
@@ -135,7 +144,7 @@ func (cc *CassandraCluster) SetDefaults() bool {
 		changed = true
 	}
 	if len(cc.Status.Phase) == 0 {
-		cc.Status.Phase = ClusterPhaseInitial
+		cc.Status.Phase = ClusterPhaseInitial.Name
 		if cc.InitCassandraRackList() < 1 {
 			logrus.Errorf("[%s]: We should have at list One Rack, Please correct the Error", cc.Name)
 		}
@@ -282,9 +291,9 @@ func (cc *CassandraCluster) initTopology(dcName string, rackName string) {
 func (cc *CassandraCluster) initCassandraRack(dcName string, rackName string) {
 	dcRackName := cc.GetDCRackName(dcName, rackName)
 	var rackStatus = CassandraRackStatus{
-		Phase: ClusterPhaseInitial,
+		Phase: ClusterPhaseInitial.Name,
 		CassandraLastAction: CassandraLastAction{
-			Name:   ClusterPhaseInitial,
+			Name:   ClusterPhaseInitial.Name,
 			Status: StatusOngoing,
 		},
 	}
@@ -300,9 +309,9 @@ func (cc *CassandraCluster) initCassandraRack(dcName string, rackName string) {
 func (cc *CassandraCluster) InitCassandraRackinStatus(status *CassandraClusterStatus, dcName string, rackName string) {
 	dcRackName := cc.GetDCRackName(dcName, rackName)
 	var rackStatus CassandraRackStatus = CassandraRackStatus{
-		Phase: ClusterPhaseInitial,
+		Phase: ClusterPhaseInitial.Name,
 		CassandraLastAction: CassandraLastAction{
-			Name:   ClusterPhaseInitial,
+			Name:   ClusterPhaseInitial.Name,
 			Status: StatusOngoing,
 		},
 	}
@@ -616,6 +625,7 @@ func (dc *DCSlice) Remove(i int) {
 func (rack *RackSlice) Remove(i int) {
 	*rack = append((*rack)[:i], (*rack)[i+1:]...)
 }
+
 // CassandraClusterSpec defines the configuration of CassandraCluster
 
 type CassandraClusterSpec struct {
@@ -727,12 +737,12 @@ type CassandraClusterSpec struct {
 // StorageConfig defines additional storage configurations
 type StorageConfig struct {
 	// Mount path into cassandra container
-	MountPath string						`json:"mountPath"`
+	MountPath string `json:"mountPath"`
 	// Name of the pvc
 	// +kubebuilder:validation:Pattern=[a-z0-9]([-a-z0-9]*[a-z0-9])?(\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*
-	Name      string						`json:"name"`
+	Name string `json:"name"`
 	// Persistent volume claim spec
-	PVCSpec   *v1.PersistentVolumeClaimSpec	`json:"pvcSpec"`
+	PVCSpec *v1.PersistentVolumeClaimSpec `json:"pvcSpec"`
 }
 
 // Topology allow to configure the Cassandra Topology according to kubernetes Nodes labels
@@ -761,7 +771,6 @@ type DC struct {
 
 	//NumTokens : configure the CASSANDRA_NUM_TOKENS parameter which can be different for each DD
 	NumTokens *int32 `json:"numTokens,omitempty"`
-
 }
 
 // Rack allow to configure Cassandra Rack according to kubernetes nodeselector labels
@@ -912,5 +921,3 @@ type CassandraClusterList struct {
 func init() {
 	SchemeBuilder.Register(&CassandraCluster{}, &CassandraClusterList{})
 }
-
-