feat: Export metrics to track number of users (#20801) (#23343)

Signed-off-by: Sathiya Narayanan Venkatesan <[email protected]>
Signed-off-by: sathiya06 <[email protected]>

Author: sathiya06

docs/operator-manual/metrics.md

@@ -210,8 +210,9 @@ argocd_cluster_labels{label_environment="production",label_team_name="team3",nam
 Metrics about API Server API request and response activity (request totals, response codes, etc...).
 Scraped at the `argocd-server-metrics:8083/metrics` endpoint.
 
-| Metric                                            |   Type    | Description                                                                                 |
-| ------------------------------------------------- | :-------: | ------------------------------------------------------------------------------------------- |
+| Metric                                            |   Type    | Description                                                                        
+|---------------------------------------------------|:---------:|---------------------------------------------------------------------------------------------|
+| `argocd_login_request_total`                      | counter   | Number of login requests.                                                                   |
 | `argocd_redis_request_duration`                   | histogram | Redis requests duration.                                                                    |
 | `argocd_redis_request_total`                      |  counter  | Number of Kubernetes requests executed during application reconciliation.                   |
 | `grpc_server_handled_total`                       |  counter  | Total number of RPCs completed on the server, regardless of success or failure.             |

server/metrics/metrics.go

@@ -20,6 +20,7 @@ type MetricsServer struct {
 	redisRequestHistogram    *prometheus.HistogramVec
 	extensionRequestCounter  *prometheus.CounterVec
 	extensionRequestDuration *prometheus.HistogramVec
+	loginRequestCounter      *prometheus.CounterVec
 }
 
 var (
@@ -53,6 +54,13 @@ var (
 		},
 		[]string{"extension"},
 	)
+	loginRequestCounter = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Name: "argocd_login_request_total",
+			Help: "Number of login requests to the Argo CD API server.",
+		},
+		[]string{"status"},
+	)
 	argoVersion = prometheus.NewGaugeVec(
 		prometheus.GaugeOpts{
 			Name: "argocd_info",
@@ -78,6 +86,7 @@ func NewMetricsServer(host string, port int) *MetricsServer {
 	registry.MustRegister(redisRequestHistogram)
 	registry.MustRegister(extensionRequestCounter)
 	registry.MustRegister(extensionRequestDuration)
+	registry.MustRegister(loginRequestCounter)
 	registry.MustRegister(argoVersion)
 
 	kubectl.RegisterWithClientGo()
@@ -92,6 +101,7 @@ func NewMetricsServer(host string, port int) *MetricsServer {
 		redisRequestHistogram:    redisRequestHistogram,
 		extensionRequestCounter:  extensionRequestCounter,
 		extensionRequestDuration: extensionRequestDuration,
+		loginRequestCounter:      loginRequestCounter,
 	}
 }
 
@@ -111,3 +121,9 @@ func (m *MetricsServer) IncExtensionRequestCounter(extension string, status int)
 func (m *MetricsServer) ObserveExtensionRequestDuration(extension string, duration time.Duration) {
 	m.extensionRequestDuration.WithLabelValues(extension).Observe(duration.Seconds())
 }
+
+// IncLoginRequestCounter increments the login request counter with the given status
+// status can be "success" or "failure"
+func (m *MetricsServer) IncLoginRequestCounter(status string) {
+	m.loginRequestCounter.WithLabelValues(status).Inc()
+}

server/server.go

@@ -573,6 +573,9 @@ func (server *ArgoCDServer) Run(ctx context.Context, listeners *Listeners) {
 	server.userStateStorage.Init(ctx)
 
 	svcSet := newArgoCDServiceSet(server)
+	if server.sessionMgr != nil {
+		server.sessionMgr.CollectMetrics(metricsServ)
+	}
 	server.serviceSet = svcSet
 	grpcS, appResourceTreeFn := server.newGRPCServer()
 	grpcWebS := grpcweb.WrapServer(grpcS)

server/session/session.go

@@ -29,6 +29,11 @@ type Authenticator interface {
 	Authenticate(ctx context.Context) (context.Context, error)
 }
 
+const (
+	success = "success"
+	failure = "failure"
+)
+
 // NewServer returns a new instance of the Session service
 func NewServer(mgr *sessionmgr.SessionManager, settingsMgr *settings.SettingsManager, authenticator Authenticator, policyEnf *rbacpolicy.RBACPolicyEnforcer, rateLimiter func() (utilio.Closer, error)) *Server {
 	return &Server{mgr, settingsMgr, authenticator, policyEnf, rateLimiter}
@@ -40,36 +45,44 @@ func (s *Server) Create(_ context.Context, q *session.SessionCreateRequest) (*se
 	if s.limitLoginAttempts != nil {
 		closer, err := s.limitLoginAttempts()
 		if err != nil {
+			s.mgr.IncLoginRequestCounter(failure)
 			return nil, err
 		}
 		defer utilio.Close(closer)
 	}
 
 	if q.Token != "" {
+		s.mgr.IncLoginRequestCounter(failure)
 		return nil, status.Errorf(codes.Unauthenticated, "token-based session creation no longer supported. please upgrade argocd cli to v0.7+")
 	}
 	if q.Username == "" || q.Password == "" {
+		s.mgr.IncLoginRequestCounter(failure)
 		return nil, status.Errorf(codes.Unauthenticated, "no credentials supplied")
 	}
 	err := s.mgr.VerifyUsernamePassword(q.Username, q.Password)
 	if err != nil {
+		s.mgr.IncLoginRequestCounter(failure)
 		return nil, err
 	}
 	uniqueId, err := uuid.NewRandom()
 	if err != nil {
+		s.mgr.IncLoginRequestCounter(failure)
 		return nil, err
 	}
 	argoCDSettings, err := s.settingsMgr.GetSettings()
 	if err != nil {
+		s.mgr.IncLoginRequestCounter(failure)
 		return nil, err
 	}
 	jwtToken, err := s.mgr.Create(
 		fmt.Sprintf("%s:%s", q.Username, settings.AccountCapabilityLogin),
 		int64(argoCDSettings.UserSessionDuration.Seconds()),
 		uniqueId.String())
 	if err != nil {
+		s.mgr.IncLoginRequestCounter(failure)
 		return nil, err
 	}
+	s.mgr.IncLoginRequestCounter(success)
 	return &session.SessionResponse{Token: jwtToken}, nil
 }
 

util/session/sessionmanager.go

@@ -43,6 +43,7 @@ type SessionManager struct {
 	sleep                         func(d time.Duration)
 	verificationDelayNoiseEnabled bool
 	failedLock                    sync.RWMutex
+	metricsRegistry               MetricsRegistry
 }
 
 // LoginAttempts is a timestamped counter for failed login attempts
@@ -53,6 +54,10 @@ type LoginAttempts struct {
 	FailCount int `json:"failCount"`
 }
 
+type MetricsRegistry interface {
+	IncLoginRequestCounter(status string)
+}
+
 const (
 	// SessionManagerClaimsIssuer fills the "iss" field of the token.
 	SessionManagerClaimsIssuer = "argocd"
@@ -172,6 +177,20 @@ func (mgr *SessionManager) Create(subject string, secondsBeforeExpiry int64, id
 	return mgr.signClaims(claims)
 }
 
+func (mgr *SessionManager) CollectMetrics(registry MetricsRegistry) {
+	mgr.metricsRegistry = registry
+	if mgr.metricsRegistry == nil {
+		log.Warn("Metrics registry is not set, metrics will not be collected")
+		return
+	}
+}
+
+func (mgr *SessionManager) IncLoginRequestCounter(status string) {
+	if mgr.metricsRegistry != nil {
+		mgr.metricsRegistry.IncLoginRequestCounter(status)
+	}
+}
+
 func (mgr *SessionManager) signClaims(claims jwt.Claims) (string, error) {
 	token := jwt.NewWithClaims(jwt.SigningMethodHS256, claims)
 	settings, err := mgr.settingsMgr.GetSettings()