Merge pull request #1685 from whitebear009/hpa-metric-type

Change the processing type from int to float in kube_horizontalpodautoscaler_spec_target_metric

Author: k8s-ci-robot

internal/store/horizontalpodautoscaler.go

@@ -36,8 +36,6 @@ const (
 	value metricTargetType = iota
 	utilization
 	average
-
-	metricTargetTypeCount // Used as a length argument to arrays
 )
 
 func (m metricTargetType) String() string {
@@ -134,53 +132,53 @@ func hpaMetricFamilies(allowAnnotationsList, allowLabelsList []string) []generat
 				for _, m := range a.Spec.Metrics {
 					var metricName string
 
-					var v [metricTargetTypeCount]int64
-					var ok [metricTargetTypeCount]bool
+					// The variable maps the type of metric to the corresponding value
+					metricMap := make(map[metricTargetType]float64)
 
 					switch m.Type {
 					case autoscaling.ObjectMetricSourceType:
 						metricName = m.Object.Metric.Name
 
-						v[value], ok[value] = m.Object.Target.Value.AsInt64()
+						if m.Object.Target.Value != nil {
+							metricMap[value] = float64(m.Object.Target.Value.MilliValue()) / 1000
+						}
 						if m.Object.Target.AverageValue != nil {
-							v[average], ok[average] = m.Object.Target.AverageValue.AsInt64()
+							metricMap[average] = float64(m.Object.Target.AverageValue.MilliValue()) / 1000
 						}
 					case autoscaling.PodsMetricSourceType:
 						metricName = m.Pods.Metric.Name
 
-						v[average], ok[average] = m.Pods.Target.AverageValue.AsInt64()
+						metricMap[average] = float64(m.Pods.Target.AverageValue.MilliValue()) / 1000
 					case autoscaling.ResourceMetricSourceType:
 						metricName = string(m.Resource.Name)
 
-						if ok[utilization] = (m.Resource.Target.AverageUtilization != nil); ok[utilization] {
-							v[utilization] = int64(*m.Resource.Target.AverageUtilization)
+						if m.Resource.Target.AverageUtilization != nil {
+							metricMap[utilization] = float64(*m.Resource.Target.AverageUtilization)
 						}
 
 						if m.Resource.Target.AverageValue != nil {
-							v[average], ok[average] = m.Resource.Target.AverageValue.AsInt64()
+							metricMap[average] = float64(m.Resource.Target.AverageValue.MilliValue()) / 1000
 						}
 					case autoscaling.ExternalMetricSourceType:
 						metricName = m.External.Metric.Name
 
 						if m.External.Target.Value != nil {
-							v[value], ok[value] = m.External.Target.Value.AsInt64()
+							metricMap[value] = float64(m.External.Target.Value.MilliValue()) / 1000
 						}
 						if m.External.Target.AverageValue != nil {
-							v[average], ok[average] = m.External.Target.AverageValue.AsInt64()
+							metricMap[average] = float64(m.External.Target.AverageValue.MilliValue()) / 1000
 						}
 					default:
 						// Skip unsupported metric type
 						continue
 					}
 
-					for i := range ok {
-						if ok[i] {
-							ms = append(ms, &metric.Metric{
-								LabelKeys:   targetMetricLabels,
-								LabelValues: []string{metricName, metricTargetType(i).String()},
-								Value:       float64(v[i]),
-							})
-						}
+					for metricTypeIndex, metricValue := range metricMap {
+						ms = append(ms, &metric.Metric{
+							LabelKeys:   targetMetricLabels,
+							LabelValues: []string{metricName, metricTypeIndex.String()},
+							Value:       metricValue,
+						})
 					}
 				}
 				return &metric.Family{Metrics: ms}

internal/store/horizontalpodautoscaler_test.go

@@ -84,6 +84,18 @@ func TestHPAStore(t *testing.T) {
 								},
 							},
 						},
+						{
+							Type: autoscaling.ObjectMetricSourceType,
+							Object: &autoscaling.ObjectMetricSource{
+								Metric: autoscaling.MetricIdentifier{
+									Name: "connections",
+								},
+								Target: autoscaling.MetricTarget{
+									Value:        resourcePtr(resource.MustParse("0.5")),
+									AverageValue: resourcePtr(resource.MustParse("0.7")),
+								},
+							},
+						},
 						{
 							Type: autoscaling.PodsMetricSourceType,
 							Pods: &autoscaling.PodsMetricSource{
@@ -195,6 +207,8 @@ func TestHPAStore(t *testing.T) {
 				kube_horizontalpodautoscaler_spec_target_metric{horizontalpodautoscaler="hpa1",metric_name="events",metric_target_type="average",namespace="ns1"} 30
 				kube_horizontalpodautoscaler_spec_target_metric{horizontalpodautoscaler="hpa1",metric_name="hits",metric_target_type="average",namespace="ns1"} 12
 				kube_horizontalpodautoscaler_spec_target_metric{horizontalpodautoscaler="hpa1",metric_name="hits",metric_target_type="value",namespace="ns1"} 10
+				kube_horizontalpodautoscaler_spec_target_metric{horizontalpodautoscaler="hpa1",metric_name="connections",metric_target_type="average",namespace="ns1"} 0.7
+				kube_horizontalpodautoscaler_spec_target_metric{horizontalpodautoscaler="hpa1",metric_name="connections",metric_target_type="value",namespace="ns1"} 0.5
 				kube_horizontalpodautoscaler_spec_target_metric{horizontalpodautoscaler="hpa1",metric_name="memory",metric_target_type="average",namespace="ns1"} 819200
 				kube_horizontalpodautoscaler_spec_target_metric{horizontalpodautoscaler="hpa1",metric_name="memory",metric_target_type="utilization",namespace="ns1"} 80
 				kube_horizontalpodautoscaler_spec_target_metric{horizontalpodautoscaler="hpa1",metric_name="sqs_jobs",metric_target_type="value",namespace="ns1"} 30