Merge branch 'ca_region_fix' of github.com:shreyas-srinivas/autoscale…

…r into ca_region_fix
kubernetes · Jun 9, 2021 · fde8ebd · fde8ebd
2 parents f6eb130 + 3fa5443
commit fde8ebd
Show file tree

Hide file tree

Showing 4 changed files with 36 additions and 30 deletions.
diff --git a/cluster-autoscaler/core/static_autoscaler.go b/cluster-autoscaler/core/static_autoscaler.go
@@ -148,12 +148,15 @@ func NewStaticAutoscaler(
 
 	scaleDown := NewScaleDown(autoscalingContext, processors, clusterStateRegistry)
 
+	// Set the initial scale times to be less than the start time so as to
+	// not start in cooldown mode.
+	initialScaleTime := time.Now().Add(-time.Hour)
 	return &StaticAutoscaler{
 		AutoscalingContext:      autoscalingContext,
 		startTime:               time.Now(),
-		lastScaleUpTime:         time.Now(),
-		lastScaleDownDeleteTime: time.Now(),
-		lastScaleDownFailTime:   time.Now(),
+		lastScaleUpTime:         initialScaleTime,
+		lastScaleDownDeleteTime: initialScaleTime,
+		lastScaleDownFailTime:   initialScaleTime,
 		scaleDown:               scaleDown,
 		processors:              processors,
 		processorCallbacks:      processorCallbacks,
@@ -257,7 +260,9 @@ func (a *StaticAutoscaler) RunOnce(currentTime time.Time) errors.AutoscalerError
 	}
 
 	// Call CloudProvider.Refresh before any other calls to cloud provider.
+	refreshStart := time.Now()
 	err = a.AutoscalingContext.CloudProvider.Refresh()
+	metrics.UpdateDurationFromStart(metrics.CloudProviderRefresh, refreshStart)
 	if err != nil {
 		klog.Errorf("Failed to refresh cloud provider config: %v", err)
 		return errors.ToAutoscalerError(errors.CloudProviderError, err)

diff --git a/cluster-autoscaler/core/static_autoscaler_test.go b/cluster-autoscaler/core/static_autoscaler_test.go
@@ -972,20 +972,20 @@ func TestStaticAutoscalerInstaceCreationErrors(t *testing.T) {
 	nodeGroupA.On("DeleteNodes", mock.Anything).Return(nil)
 	nodeGroupA.On("Nodes").Return([]cloudprovider.Instance{
 		{
-			"A1",
-			&cloudprovider.InstanceStatus{
+			Id: "A1",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceRunning,
 			},
 		},
 		{
-			"A2",
-			&cloudprovider.InstanceStatus{
+			Id: "A2",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceCreating,
 			},
 		},
 		{
-			"A3",
-			&cloudprovider.InstanceStatus{
+			Id: "A3",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceCreating,
 				ErrorInfo: &cloudprovider.InstanceErrorInfo{
 					ErrorClass: cloudprovider.OutOfResourcesErrorClass,
@@ -994,8 +994,8 @@ func TestStaticAutoscalerInstaceCreationErrors(t *testing.T) {
 			},
 		},
 		{
-			"A4",
-			&cloudprovider.InstanceStatus{
+			Id: "A4",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceCreating,
 				ErrorInfo: &cloudprovider.InstanceErrorInfo{
 					ErrorClass: cloudprovider.OutOfResourcesErrorClass,
@@ -1004,8 +1004,8 @@ func TestStaticAutoscalerInstaceCreationErrors(t *testing.T) {
 			},
 		},
 		{
-			"A5",
-			&cloudprovider.InstanceStatus{
+			Id: "A5",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceCreating,
 				ErrorInfo: &cloudprovider.InstanceErrorInfo{
 					ErrorClass: cloudprovider.OutOfResourcesErrorClass,
@@ -1014,8 +1014,8 @@ func TestStaticAutoscalerInstaceCreationErrors(t *testing.T) {
 			},
 		},
 		{
-			"A6",
-			&cloudprovider.InstanceStatus{
+			Id: "A6",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceCreating,
 				ErrorInfo: &cloudprovider.InstanceErrorInfo{
 					ErrorClass: cloudprovider.OtherErrorClass,
@@ -1032,8 +1032,8 @@ func TestStaticAutoscalerInstaceCreationErrors(t *testing.T) {
 	nodeGroupB.On("DeleteNodes", mock.Anything).Return(nil)
 	nodeGroupB.On("Nodes").Return([]cloudprovider.Instance{
 		{
-			"B1",
-			&cloudprovider.InstanceStatus{
+			Id: "B1",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceRunning,
 			},
 		},
@@ -1102,38 +1102,38 @@ func TestStaticAutoscalerInstaceCreationErrors(t *testing.T) {
 	// restub node group A so nodes are no longer reporting errors
 	nodeGroupA.On("Nodes").Return([]cloudprovider.Instance{
 		{
-			"A1",
-			&cloudprovider.InstanceStatus{
+			Id: "A1",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceRunning,
 			},
 		},
 		{
-			"A2",
-			&cloudprovider.InstanceStatus{
+			Id: "A2",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceCreating,
 			},
 		},
 		{
-			"A3",
-			&cloudprovider.InstanceStatus{
+			Id: "A3",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceDeleting,
 			},
 		},
 		{
-			"A4",
-			&cloudprovider.InstanceStatus{
+			Id: "A4",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceDeleting,
 			},
 		},
 		{
-			"A5",
-			&cloudprovider.InstanceStatus{
+			Id: "A5",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceDeleting,
 			},
 		},
 		{
-			"A6",
-			&cloudprovider.InstanceStatus{
+			Id: "A6",
+			Status: &cloudprovider.InstanceStatus{
 				State: cloudprovider.InstanceDeleting,
 			},
 		},

diff --git a/cluster-autoscaler/metrics/metrics.go b/cluster-autoscaler/metrics/metrics.go
@@ -90,6 +90,7 @@ const (
 	FindUnneeded               FunctionLabel = "findUnneeded"
 	UpdateState                FunctionLabel = "updateClusterState"
 	FilterOutSchedulable       FunctionLabel = "filterOutSchedulable"
+	CloudProviderRefresh       FunctionLabel = "cloudProviderRefresh"
 	Main                       FunctionLabel = "main"
 	Poll                       FunctionLabel = "poll"
 	Reconfigure                FunctionLabel = "reconfigure"

diff --git a/cluster-autoscaler/simulator/cluster_test.go b/cluster-autoscaler/simulator/cluster_test.go
@@ -69,7 +69,7 @@ func TestUtilization(t *testing.T) {
 	assert.InEpsilon(t, 2.0/10, utilInfo.Utilization, 0.01)
 
 	terminatedPod := BuildTestPod("podTerminated", 100, 200000)
-	terminatedPod.DeletionTimestamp = &metav1.Time{testTime.Add(-10 * time.Minute)}
+	terminatedPod.DeletionTimestamp = &metav1.Time{Time: testTime.Add(-10 * time.Minute)}
 	nodeInfo = schedulerframework.NewNodeInfo(pod, pod, pod2, terminatedPod)
 	utilInfo, err = CalculateUtilization(node, nodeInfo, false, false, gpuLabel, testTime)
 	assert.NoError(t, err)