Add pprof endpoint flag to the operator

cmd/gpu-operator/main.go

@@ -78,9 +78,12 @@ func main() {
 	var enableLeaderElection bool
 	var leaderElectionNamespace string
 	var probeAddr string
+	var pprofAddr string
 	var renewDeadline time.Duration
 
 	flag.StringVar(&metricsAddr, "metrics-bind-address", ":8080", "The address the metric endpoint binds to.")
+	flag.StringVar(&pprofAddr, "pprof-bind-address", "",
+		"The address the pprof endpoint binds to (e.g. \":6060\"). Disabled when empty.")
 	flag.StringVar(&probeAddr, "health-probe-bind-address", ":8081", "The address the probe endpoint binds to.")
 	flag.BoolVar(&enableLeaderElection, "leader-elect", false,
 		"Enable leader election for controller manager. "+
@@ -136,6 +139,7 @@ func main() {
 		Scheme:                  scheme,
 		Metrics:                 metricsOptions,
 		HealthProbeBindAddress:  probeAddr,
+		PprofBindAddress:        pprofAddr,
 		LeaderElection:          enableLeaderElection,
 		LeaderElectionNamespace: leaderElectionNamespace,
 		LeaderElectionID:        "53822513.nvidia.com",

deployments/gpu-operator/templates/operator.yaml

@@ -49,6 +49,9 @@ spec:
         {{- if .Values.operator.logging.level }}
         - --zap-log-level={{- .Values.operator.logging.level }}
         {{- end }}
+      {{- end }}
+      {{- if .Values.operator.pprof.bindAddress }}
+        - --pprof-bind-address={{ .Values.operator.pprof.bindAddress }}
       {{- end }}
         env:
         - name: WATCH_NAMESPACE

deployments/gpu-operator/values.yaml

@@ -111,6 +111,9 @@ operator:
     # Development Mode defaults(encoder=consoleEncoder,logLevel=Debug,stackTraceLevel=Warn)
     # Production Mode defaults(encoder=jsonEncoder,logLevel=Info,stackTraceLevel=Error)
     develMode: false
+  pprof:
+    # Bind address for the pprof profiling endpoint (e.g. ":6060"). Disabled when empty.
+    bindAddress: ""
   resources:
     limits:
       cpu: 500m