fix(server): re-enable prometheus counters

Commit: 0213c08 ("tracee: add ready callback") made prometheus stats to
stop being registered due to a timing issue. This patch re-enables it
by registering prometheus counters in the callback function itself
instead of in Tracee.New() (too early).

Fixes: #3303

pkg/cmd/flags/server/server.go

@@ -2,6 +2,7 @@ package server
 
 import (
 	"github.com/aquasecurity/tracee/pkg/errfmt"
+	"github.com/aquasecurity/tracee/pkg/logger"
 	"github.com/aquasecurity/tracee/pkg/server"
 )
 
@@ -28,17 +29,21 @@ func PrepareServer(listenAddr string, metrics, healthz, pprof, pyro bool) (*serv
 		httpServer := server.New(listenAddr)
 
 		if metrics {
+			logger.Debugw("Enabling metrics endpoint")
 			httpServer.EnableMetricsEndpoint()
 		}
 
 		if healthz {
+			logger.Debugw("Enabling healthz endpoint")
 			httpServer.EnableHealthzEndpoint()
 		}
 
 		if pprof {
+			logger.Debugw("Enabling pprof endpoint")
 			httpServer.EnablePProfEndpoint()
 		}
 		if pyro {
+			logger.Debugw("Enabling pyroscope agent")
 			err := httpServer.EnablePyroAgent()
 			if err != nil {
 				return httpServer, err

pkg/cmd/tracee.go

@@ -32,16 +32,23 @@ func (r Runner) Run(ctx context.Context) error {
 		return errfmt.Errorf("error creating Tracee: %v", err)
 	}
 
-	// start the server and write pid file only after tracee is ready
-	t.AddReadyCallback(func(ctx context.Context) {
-		logger.Debugw("Tracee is ready callback")
+	// Readiness Callback: Tracee is ready to receive events
 
-		// start server if one is configured
-		if r.Server != nil {
-			r.TraceeConfig.MetricsEnabled = r.Server.MetricsEndpointEnabled()
+	t.AddReadyCallback(
+		func(ctx context.Context) {
+			logger.Debugw("Tracee is ready callback")
+			if r.Server == nil {
+				return
+			}
+			if r.Server.MetricsEndpointEnabled() {
+				r.TraceeConfig.MetricsEnabled = true // TODO: is this needed ?
+				if err := t.Stats().RegisterPrometheus(); err != nil {
+					logger.Errorw("Registering prometheus metrics", "error", err)
+				}
+			}
 			go r.Server.Start(ctx)
-		}
-	})
+		},
+	)
 
 	// Initialize tracee
 
@@ -51,7 +58,7 @@ func (r Runner) Run(ctx context.Context) error {
 	}
 
 	// Manage PID file
-	// write pid file
+
 	if err := writePidFile(t.OutDir); err != nil {
 		return errfmt.WrapError(err)
 	}
@@ -61,11 +68,12 @@ func (r Runner) Run(ctx context.Context) error {
 		}
 	}()
 
-	// preeamble
+	// Preeamble
 
 	r.Printer.Preamble()
 
 	// Start event channel reception
+
 	go func() {
 		for {
 			select {
@@ -77,16 +85,17 @@ func (r Runner) Run(ctx context.Context) error {
 		}
 	}()
 
-	// This will block until the context is done
+	// Blocks (until ctx is Done)
+
 	err = t.Run(ctx)
 
-	// Drain and print the remaining channel events that were sent before full termination
+	// Drain remaininig channel events (sent during shutdown)
+
 	for {
 		select {
 		case event := <-r.TraceeConfig.ChanEvents:
 			r.Printer.Print(event)
 		default:
-			// Print statistics at the end
 			stats := t.Stats()
 			r.Printer.Epilogue(*stats)
 			r.Printer.Close()

pkg/ebpf/tracee.go

@@ -307,15 +307,6 @@ func New(cfg config.Config) (*Tracee, error) {
 
 	t.triggerContexts = trigger.NewContext()
 
-	// Export metrics to prometheus if enabled
-
-	if t.config.MetricsEnabled {
-		err := t.Stats().RegisterPrometheus()
-		if err != nil {
-			logger.Errorw("Registering prometheus metrics", "error", err)
-		}
-	}
-
 	return t, nil
 }