Merge pull request #115082 from sanposhiho/filter-metrics

feature(scheduler): implement plugin_evaluation_total metric

pkg/scheduler/framework/runtime/framework.go

@@ -617,6 +617,7 @@ func (f *frameworkImpl) RunPreFilterPlugins(ctx context.Context, state *framewor
 			skipPlugins.Insert(pl.Name())
 			continue
 		}
+		metrics.PluginEvaluationTotal.WithLabelValues(pl.Name(), preFilter, f.profileName).Inc()
 		if !s.IsSuccess() {
 			s.SetFailedPlugin(pl.Name())
 			if s.IsUnschedulable() {
@@ -734,6 +735,7 @@ func (f *frameworkImpl) RunFilterPlugins(
 		if state.SkipFilterPlugins.Has(pl.Name()) {
 			continue
 		}
+		metrics.PluginEvaluationTotal.WithLabelValues(pl.Name(), Filter, f.profileName).Inc()
 		if status := f.runFilterPlugin(ctx, pl, state, pod, nodeInfo); !status.IsSuccess() {
 			if !status.IsUnschedulable() {
 				// Filter plugins are not supposed to return any status other than

pkg/scheduler/metrics/metrics.go

@@ -189,6 +189,14 @@ var (
 			StabilityLevel: metrics.ALPHA,
 		}, []string{"plugin", "profile"})
 
+	PluginEvaluationTotal = metrics.NewCounterVec(
+		&metrics.CounterOpts{
+			Subsystem:      SchedulerSubsystem,
+			Name:           "plugin_evaluation_total",
+			Help:           "Number of attempts to schedule pods by each plugin and the extension point (available only in PreFilter and Filter.).",
+			StabilityLevel: metrics.ALPHA,
+		}, []string{"plugin", "extension_point", "profile"})
+
 	metricsList = []metrics.Registerable{
 		scheduleAttempts,
 		e2eSchedulingLatency,
@@ -207,6 +215,7 @@ var (
 		PermitWaitDuration,
 		CacheSize,
 		unschedulableReasons,
+		PluginEvaluationTotal,
 	}
 )