test(prof): Profiler handlers for debugging support

cmd/node-observer/main.go

@@ -83,5 +83,15 @@ func mainInternal(c string) error {
 	// Controller
 	g.Add(controller.Start, controller.Stop)
 
+	//Register profiling endpoinds if enabled
+	//if cfg.EnableProfiling {
+	profilingPort := cfg.ProfilingPort
+	if profilingPort == 0 {
+		profilingPort = node_observer.DefaultProfilingPort
+	}
+	profilingServer := node_observer.NewProfilingServer(profilingPort)
+	g.Add(profilingServer.Start, profilingServer.Stop)
+	//}
+
 	return g.Run()
 }

pkg/config/config.go

@@ -38,6 +38,7 @@ type Config struct {
 	CredsPath               *string           `yaml:"credentialsPath,omitempty"`
 	FwdSvcURL               *string           `yaml:"forwardServiceUrl,omitempty"`
 	Env                     map[string]string `yaml:"env"`
+	EnableProfiling         bool              `yaml:"enableProfiling,omitempty"`
 
 	// derived
 	Credentials map[string]any

pkg/node_observer/config.go

@@ -26,12 +26,16 @@ import (
 	"github.com/NVIDIA/topograph/pkg/topology"
 )
 
+const DefaultProfilingPort = 49021
+
 type Config struct {
 	GenerateTopologyURL string            `yaml:"generateTopologyUrl"`
 	Trigger             Trigger           `yaml:"trigger"`
 	Provider            topology.Provider `yaml:"provider"`
 	Engine              topology.Engine   `yaml:"engine"`
 	Params              map[string]any    `yaml:"params"`
+	EnableProfiling     bool              `yaml:"enableProfiling,omitempty"`
+	ProfilingPort       int               `yaml:"profilingPort,omitempty"`
 }
 
 type Trigger struct {

pkg/node_observer/profiler.go

@@ -0,0 +1,43 @@
+package node_observer
+
+import (
+	"fmt"
+	"net"
+	"net/http"
+	_ "net/http/pprof"
+
+	"k8s.io/klog/v2"
+)
+
+type Profiler struct {
+	listener net.Listener
+}
+
+func NewProfilingServer(port int) *Profiler {
+	// Listen on the specified port for pprof profiling
+	addr := net.JoinHostPort("localhost", fmt.Sprintf("%d", port))
+	listener, err := net.Listen("tcp", addr)
+	if err != nil {
+		klog.Fatalf("Failed to start profiling server on %s: %v", addr, err)
+	}
+
+	return &Profiler{
+		listener: listener,
+	}
+}
+
+func (c *Profiler) Start() error {
+	// Start the pprof server
+	err := http.Serve(c.listener, nil) // DefaultServeMux will handle pprof endpoints
+	if err != nil {
+		klog.Errorf("Failed to start pprof server: %v", err)
+		return err
+	}
+	klog.Infof("Pprofiler server started on %s", c.listener.Addr().String())
+	return nil
+}
+
+func (c *Profiler) Stop(err error) {
+	klog.Infof("Stopping Pprofiler server: %v", err)
+	c.listener.Close()
+}

pkg/server/http_server.go

@@ -22,6 +22,7 @@ import (
 	"io"
 	"net"
 	"net/http"
+	"net/http/pprof"
 	"time"
 
 	"github.com/prometheus/client_golang/prometheus/promhttp"
@@ -99,6 +100,16 @@ func initHttpServer(ctx context.Context, cfg *config.Config) *HttpServer {
 	mux.HandleFunc("/healthz", healthz)
 	mux.Handle("/metrics", promhttp.Handler())
 
+	// Register pprof handlers for performance profiling
+	//if cfg.EnableProfiling {
+	klog.Infof("Enabling pprof profiling endpoints")
+	mux.HandleFunc("/debug/pprof/", pprof.Index)
+	mux.HandleFunc("/debug/pprof/cmdline", pprof.Cmdline)
+	mux.HandleFunc("/debug/pprof/profile", pprof.Profile)
+	mux.HandleFunc("/debug/pprof/symbol", pprof.Symbol)
+	mux.HandleFunc("/debug/pprof/trace", pprof.Trace)
+	//}
+
 	return &HttpServer{
 		ctx: ctx,
 		cfg: cfg,