adds flag to disble pprof if desired, defaulted on

cmd/epp/runner/runner.go

@@ -65,6 +65,10 @@ var (
 		"metricsPort",
 		runserver.DefaultMetricsPort,
 		"The metrics port")
+	enablePprof = flag.Bool(
+		"enablePprof",
+		runserver.DefaultEnablePprof,
+		"Enables pprof handlers. Defaults to true. Set to false to disable pprof handlers.")
 	destinationEndpointHintKey = flag.String(
 		"destinationEndpointHintKey",
 		runserver.DefaultDestinationEndpointHintKey,
@@ -263,10 +267,14 @@ func (r *Runner) Run(ctx context.Context) error {
 		setupLog.Error(err, "Failed to create controller manager")
 		return err
 	}
-	err = setupPprofHandlers(mgr)
-	if err != nil {
-		setupLog.Error(err, "Failed to setup pprof handlers")
-		return err
+
+	if *enablePprof {
+		setupLog.Info("Enabling pprof handlers")
+		err = setupPprofHandlers(mgr)
+		if err != nil {
+			setupLog.Error(err, "Failed to setup pprof handlers")
+			return err
+		}
 	}
 
 	err = r.parseConfiguration(ctx)

config/charts/inferencepool/templates/epp-deployment.yaml

@@ -35,6 +35,8 @@ spec:
         - "9003"
         - -metricsPort
         - "9090"
+        # https://pkg.go.dev/flag#hdr-Command_line_flag_syntax; space is only for non-bool flags
+        - "-enablePprof={{ .Values.inferenceExtension.enablePprof }}"
         {{- if eq (.Values.inferencePool.modelServerType | default "vllm") "triton-tensorrt-llm" }}
         - -totalQueuedRequestsMetric
         - "nv_trt_llm_request_metrics{request_type=waiting}"

config/charts/inferencepool/values.yaml

@@ -7,6 +7,7 @@ inferenceExtension:
     pullPolicy: Always
   extProcPort: 9002
   env: {}
+  enablePprof: true # Enable pprof handlers for profiling and debugging
   # Example environment variables:
   # env:
   #   KV_CACHE_SCORE_WEIGHT: "1"

pkg/epp/server/runserver.go

@@ -72,6 +72,7 @@ const (
 	DefaultRefreshPrometheusMetricsInterval         = 5 * time.Second                  // default for --refreshPrometheusMetricsInterval
 	DefaultSecureServing                            = true                             // default for --secureServing
 	DefaultHealthChecking                           = false                            // default for --healthChecking
+	DefaultEnablePprof                              = true                             // default for --enablePprof
 	DefaultTotalQueuedRequestsMetric                = "vllm:num_requests_waiting"      // default for --totalQueuedRequestsMetric
 	DefaultKvCacheUsagePercentageMetric             = "vllm:gpu_cache_usage_perc"      // default for --kvCacheUsagePercentageMetric
 	DefaultLoraInfoMetric                           = "vllm:lora_requests_info"        // default for --loraInfoMetric