Merge pull request #72140 from wgliang/feature/adaptive-percentage-find-nodes

Set percentage of nodes scored in each cycle dynamically based on the…

cmd/kube-scheduler/app/options/options_test.go

@@ -225,8 +225,7 @@ users:
 					Burst:       100,
 					ContentType: "application/vnd.kubernetes.protobuf",
 				},
-				PercentageOfNodesToScore: 50,
+				BindTimeoutSeconds: &defaultBindTimeoutSeconds,
-				BindTimeoutSeconds:       &defaultBindTimeoutSeconds,
 			},
 		},
 		{
@@ -306,8 +305,7 @@ users:
 					Burst:       100,
 					ContentType: "application/vnd.kubernetes.protobuf",
 				},
-				PercentageOfNodesToScore: 50,
+				BindTimeoutSeconds: &defaultBindTimeoutSeconds,
-				BindTimeoutSeconds:       &defaultBindTimeoutSeconds,
 			},
 		},
 		{

pkg/scheduler/apis/config/types.go

@@ -79,7 +79,8 @@ type KubeSchedulerConfiguration struct {
 	// at least "minFeasibleNodesToFind" feasible nodes no matter what the value of this flag is.
 	// Example: if the cluster size is 500 nodes and the value of this flag is 30,
 	// then scheduler stops finding further feasible nodes once it finds 150 feasible ones.
-	// When the value is 0, default percentage (50%) of the nodes will be scored.
+	// When the value is 0, default percentage (5%--50% based on the size of the cluster) of the
+	// nodes will be scored.
 	PercentageOfNodesToScore int32
 
 	// DEPRECATED.

pkg/scheduler/apis/config/v1alpha1/defaults.go

@@ -81,11 +81,6 @@ func SetDefaults_KubeSchedulerConfiguration(obj *kubescedulerconfigv1alpha1.Kube
 		obj.LeaderElection.LockObjectName = kubescedulerconfigv1alpha1.SchedulerDefaultLockObjectName
 	}
 
-	if obj.PercentageOfNodesToScore == 0 {
-		// by default, stop finding feasible nodes once the number of feasible nodes is 50% of the cluster.
-		obj.PercentageOfNodesToScore = 50
-	}
-
 	if len(obj.FailureDomains) == 0 {
 		obj.FailureDomains = kubeletapis.DefaultFailureDomains
 	}

pkg/scheduler/core/generic_scheduler.go

@@ -54,6 +54,11 @@ const (
 	// certain minimum of nodes are checked for feasibility. This in turn helps
 	// ensure a minimum level of spreading.
 	minFeasibleNodesToFind = 100
+	// minFeasibleNodesPercentageToFind is the minimum percentage of nodes that
+	// would be scored in each scheduling cycle. This is a semi-arbitrary value
+	// to ensure that a certain minimum of nodes are checked for feasibility.
+	// This in turn helps ensure a minimum level of spreading.
+	minFeasibleNodesPercentageToFind = 5
 )
 
 // FailedPredicateMap declares a map[string][]algorithm.PredicateFailureReason type.
@@ -375,15 +380,24 @@ func (g *genericScheduler) getLowerPriorityNominatedPods(pod *v1.Pod, nodeName s
 
 // numFeasibleNodesToFind returns the number of feasible nodes that once found, the scheduler stops
 // its search for more feasible nodes.
-func (g *genericScheduler) numFeasibleNodesToFind(numAllNodes int32) int32 {
+func (g *genericScheduler) numFeasibleNodesToFind(numAllNodes int32) (numNodes int32) {
-	if numAllNodes < minFeasibleNodesToFind || g.percentageOfNodesToScore <= 0 ||
+	if numAllNodes < minFeasibleNodesToFind || g.percentageOfNodesToScore >= 100 {
-		g.percentageOfNodesToScore >= 100 {
 		return numAllNodes
 	}
-	numNodes := numAllNodes * g.percentageOfNodesToScore / 100
+
+	adaptivePercentage := g.percentageOfNodesToScore
+	if adaptivePercentage <= 0 {
+		adaptivePercentage = schedulerapi.DefaultPercentageOfNodesToScore - numAllNodes/125
+		if adaptivePercentage < minFeasibleNodesPercentageToFind {
+			adaptivePercentage = minFeasibleNodesPercentageToFind
+		}
+	}
+
+	numNodes = numAllNodes * adaptivePercentage / 100
 	if numNodes < minFeasibleNodesToFind {
 		return minFeasibleNodesToFind
 	}
+
 	return numNodes
 }
 

pkg/scheduler/core/generic_scheduler_test.go

@@ -1483,3 +1483,56 @@ func TestPreempt(t *testing.T) {
 		})
 	}
 }
+
+func TestNumFeasibleNodesToFind(t *testing.T) {
+	tests := []struct {
+		name                     string
+		percentageOfNodesToScore int32
+		numAllNodes              int32
+		wantNumNodes             int32
+	}{
+		{
+			name:         "not set percentageOfNodesToScore and nodes number not more than 50",
+			numAllNodes:  10,
+			wantNumNodes: 10,
+		},
+		{
+			name:                     "set percentageOfNodesToScore and nodes number not more than 50",
+			percentageOfNodesToScore: 40,
+			numAllNodes:              10,
+			wantNumNodes:             10,
+		},
+		{
+			name:         "not set percentageOfNodesToScore and nodes number more than 50",
+			numAllNodes:  1000,
+			wantNumNodes: 420,
+		},
+		{
+			name:                     "set percentageOfNodesToScore and nodes number more than 50",
+			percentageOfNodesToScore: 40,
+			numAllNodes:              1000,
+			wantNumNodes:             400,
+		},
+		{
+			name:         "not set percentageOfNodesToScore and nodes number more than 50*125",
+			numAllNodes:  6000,
+			wantNumNodes: 300,
+		},
+		{
+			name:                     "set percentageOfNodesToScore and nodes number more than 50*125",
+			percentageOfNodesToScore: 40,
+			numAllNodes:              6000,
+			wantNumNodes:             2400,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			g := &genericScheduler{
+				percentageOfNodesToScore: tt.percentageOfNodesToScore,
+			}
+			if gotNumNodes := g.numFeasibleNodesToFind(tt.numAllNodes); gotNumNodes != tt.wantNumNodes {
+				t.Errorf("genericScheduler.numFeasibleNodesToFind() = %v, want %v", gotNumNodes, tt.wantNumNodes)
+			}
+		})
+	}
+}

staging/src/k8s.io/kube-scheduler/config/v1alpha1/types.go

@@ -75,7 +75,8 @@ type KubeSchedulerConfiguration struct {
 	// at least "minFeasibleNodesToFind" feasible nodes no matter what the value of this flag is.
 	// Example: if the cluster size is 500 nodes and the value of this flag is 30,
 	// then scheduler stops finding further feasible nodes once it finds 150 feasible ones.
-	// When the value is 0, default percentage (50%) of the nodes will be scored.
+	// When the value is 0, default percentage (5%--50% based on the size of the cluster) of the
+	// nodes will be scored.
 	PercentageOfNodesToScore int32 `json:"percentageOfNodesToScore"`
 
 	// DEPRECATED.