koord-scheduler: support log filtering failure reasons as needed

Signed-off-by: wangjianyu <zmsjianyu@gmail.com>

Author: ZiMengSheng

cmd/koord-scheduler/app/server.go

@@ -273,6 +273,7 @@ func installProfilingHandler(pathRecorderMux *mux.PathRecorderMux, enableContent
 	debugFlags := utilroutes.NewDebugFlags(pathRecorderMux)
 	debugFlags.Install("v", utilroutes.StringFlagPutHandler(logs.GlogSetter))
 	debugFlags.Install("s", utilroutes.StringFlagPutHandler(frameworkext.DebugScoresSetter))
+	debugFlags.Install("f", utilroutes.StringFlagPutHandler(frameworkext.DebugFiltersSetter))
 }
 
 // newMetricsHandler builds a metrics server from the config.

…g/scheduler/frameworkext/debug_scores.go pkg/scheduler/frameworkext/debug.gopkg/scheduler/frameworkext/debug_scores.go renamed to pkg/scheduler/frameworkext/debug.go pkg/scheduler/frameworkext/debug_scores.go renamed to pkg/scheduler/frameworkext/debug.go

@@ -29,11 +29,13 @@ import (
 )
 
 var (
-	debugTopNScores = 0
+	debugTopNScores    = 0
+	debugFilterFailure = false
 )
 
 func AddFlags(fs *pflag.FlagSet) {
 	fs.IntVarP(&debugTopNScores, "debug-scores", "s", debugTopNScores, "logging topN nodes score and scores for each plugin after running the score extension, disable if set to 0")
+	fs.BoolVarP(&debugFilterFailure, "debug-filters", "f", debugFilterFailure, "logging filter failures")
 }
 
 // DebugScoresSetter updates debugTopNScores to specified value
@@ -46,6 +48,16 @@ func DebugScoresSetter(val string) (string, error) {
 	return fmt.Sprintf("successfully set debugTopNScores to %s", val), nil
 }
 
+// DebugFiltersSetter updates debugFilterFailure to specified value
+func DebugFiltersSetter(val string) (string, error) {
+	filterFailure, err := strconv.ParseBool(val)
+	if err != nil {
+		return "", fmt.Errorf("failed set debugFilterFailure %s: %v", val, err)
+	}
+	debugFilterFailure = filterFailure
+	return fmt.Sprintf("successfully set debugFilterFailure to %s", val), nil
+}
+
 func debugScores(topN int, pod *corev1.Pod, pluginToNodeScores map[string]framework.NodeScoreList, nodes []*corev1.Node) prettytable.Writer {
 	// Summarize all scores.
 	result := make(framework.NodeScoreList, 0, len(nodes))

…eduler/frameworkext/debug_scores_test.go pkg/scheduler/frameworkext/debug_test.gopkg/scheduler/frameworkext/debug_scores_test.go renamed to pkg/scheduler/frameworkext/debug_test.go pkg/scheduler/frameworkext/debug_scores_test.go renamed to pkg/scheduler/frameworkext/debug_test.go

@@ -54,6 +54,40 @@ func TestDebugScoresSetter(t *testing.T) {
 	}
 }
 
+func TestDebugFiltersSetter(t *testing.T) {
+	tests := []struct {
+		name    string
+		value   string
+		wantErr bool
+		want    bool
+	}{
+		{
+			name:    "valid bool",
+			value:   "true",
+			wantErr: false,
+			want:    true,
+		},
+		{
+			name:    "invalid",
+			value:   "11.22",
+			wantErr: true,
+			want:    false,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			_, err := DebugFiltersSetter(tt.value)
+			if tt.wantErr && err == nil {
+				t.Error("expected error but got nil")
+			} else if !tt.wantErr && err != nil {
+				t.Errorf("expected no error but got err: %v", err)
+			}
+			assert.Equal(t, tt.want, debugFilterFailure)
+			debugFilterFailure = false
+		})
+	}
+}
+
 func TestDebugScores(t *testing.T) {
 	pluginToNodeScores := map[string]framework.NodeScoreList{
 		"ImageLocality": {

pkg/scheduler/frameworkext/framework_extender.go

@@ -227,7 +227,11 @@ func (ext *frameworkExtenderImpl) RunFilterPluginsWithNominatedPods(ctx context.
 			nodeInfo = newNodeInfo
 		}
 	}
-	return ext.Framework.RunFilterPluginsWithNominatedPods(ctx, cycleState, pod, nodeInfo)
+	status := ext.Framework.RunFilterPluginsWithNominatedPods(ctx, cycleState, pod, nodeInfo)
+	if !status.IsSuccess() && debugFilterFailure {
+		klog.Infof("Failed to filter for Pod %q on Node %q, failedPlugin: %s, reason: %s", klog.KObj(pod), klog.KObj(nodeInfo.Node()), status.FailedPlugin(), status.Message())
+	}
+	return status
 }
 
 func (ext *frameworkExtenderImpl) RunScorePlugins(ctx context.Context, state *framework.CycleState, pod *corev1.Pod, nodes []*corev1.Node) (framework.PluginToNodeScores, *framework.Status) {