add metric ext_cpu_total_distribute for local state

examples/ensurance/disablescheduling-when-ext-cpu-total-distribute.yaml

@@ -0,0 +1,20 @@
+apiVersion: ensurance.crane.io/v1alpha1
+kind: NodeQOSEnsurancePolicy
+metadata:
+  name: "disablescheduling-when-ext-cpu-total-distribute"
+  labels:
+    app: "system"
+spec:
+  nodeQualityProbe:
+    timeoutSeconds: 10
+    nodeLocalGet:
+      localCacheTTLSeconds: 60
+  objectiveEnsurances:
+    - name: "ext_cpu_total_distribute"
+      avoidanceThreshold: 2
+      restoreThreshold: 2
+      actionName: "disablescheduling"
+      strategy: "None"
+      metricRule:
+        name: "ext_cpu_total_distribute"
+        value: 110

pkg/ensurance/collector/collector.go

@@ -14,6 +14,7 @@ import (
 	"github.com/gocrane/crane/pkg/common"
 	"github.com/gocrane/crane/pkg/ensurance/collector/cadvisor"
 	"github.com/gocrane/crane/pkg/ensurance/collector/nodelocal"
+	"github.com/gocrane/crane/pkg/ensurance/collector/noderesource"
 	"github.com/gocrane/crane/pkg/ensurance/collector/types"
 	"github.com/gocrane/crane/pkg/features"
 	"github.com/gocrane/crane/pkg/known"
@@ -179,6 +180,15 @@ func (s *StateCollector) UpdateCollectors() {
 		if _, exists := s.collectors.Load(types.CadvisorCollectorType); !exists {
 			s.collectors.Store(types.CadvisorCollectorType, cadvisor.NewCadvisorCollector(s.podLister, s.GetCadvisorManager()))
 		}
+
+		if nodeResourceGate := utilfeature.DefaultFeatureGate.Enabled(features.CraneNodeResource); nodeResourceGate {
+			if _, exists := s.collectors.Load(types.NodeResourceCollectorType); !exists {
+				c := noderesource.NewNodeResourceCollector(s.nodeName, s.nodeLister, s.podLister)
+				if c != nil {
+					s.collectors.Store(types.NodeResourceCollectorType, c)
+				}
+			}
+		}
 		break
 	}
 

pkg/ensurance/collector/noderesource/noderesource.go

@@ -0,0 +1,63 @@
+package noderesource
+
+import (
+	"fmt"
+	"time"
+
+	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
+	"k8s.io/apimachinery/pkg/labels"
+	v1 "k8s.io/client-go/listers/core/v1"
+	"k8s.io/klog/v2"
+
+	"github.com/gocrane/crane/pkg/common"
+	"github.com/gocrane/crane/pkg/ensurance/collector/types"
+	"github.com/gocrane/crane/pkg/utils"
+)
+
+type NodeResource struct {
+	nodeName   string
+	nodeLister v1.NodeLister
+	podLister  v1.PodLister
+}
+
+func NewNodeResourceCollector(nodeName string, nodeLister v1.NodeLister, podLister v1.PodLister) *NodeResource {
+	klog.V(4).Infof("NodeResourceCollector create")
+	return &NodeResource{
+		nodeName:   nodeName,
+		nodeLister: nodeLister,
+		podLister:  podLister,
+	}
+}
+
+func (n *NodeResource) GetType() types.CollectType {
+	return types.NodeResourceCollectorType
+}
+
+func (n *NodeResource) Collect() (map[string][]common.TimeSeries, error) {
+	klog.V(6).Infof("NodeResourceCollector Collect")
+	node, err := n.nodeLister.Get(n.nodeName)
+	if err != nil {
+		return nil, err
+	}
+	pods, err := n.podLister.List(labels.Everything())
+	if err != nil {
+		return nil, err
+	}
+
+	allExtCpu := node.Status.Allocatable.Name(corev1.ResourceName(fmt.Sprintf(utils.ExtResourcePrefixFormat, corev1.ResourceCPU.String())), resource.DecimalSI).MilliValue()
+	var distributeExtCpu int64 = 0
+	for _, pod := range pods {
+		for _, container := range pod.Spec.Containers {
+			if quantity, ok := container.Resources.Requests[corev1.ResourceName(fmt.Sprintf(utils.ExtResourcePrefixFormat, corev1.ResourceCPU.String()))]; ok {
+				distributeExtCpu += quantity.MilliValue()
+			}
+		}
+	}
+	klog.V(4).Infof("allExtCpu: %d, distributeExtCpu: %d", allExtCpu, distributeExtCpu)
+	return map[string][]common.TimeSeries{string(types.MetricNameExtCpuTotalDistribute): {{Samples: []common.Sample{{Value: (float64(distributeExtCpu) / float64(allExtCpu)) * 100, Timestamp: time.Now().Unix()}}}}}, nil
+}
+
+func (n *NodeResource) Stop() error {
+	return nil
+}

pkg/ensurance/collector/types/types.go

@@ -15,6 +15,7 @@ const (
 	CadvisorCollectorType      CollectType = "cadvisor"
 	EbpfCollectorType          CollectType = "ebpf"
 	MetricsServerCollectorType CollectType = "metrics-server"
+	NodeResourceCollectorType  CollectType = "node-resource"
 )
 
 type MetricName string
@@ -51,6 +52,7 @@ const (
 	MetricNameContainerSchedRunQueueTime MetricName = "container_sched_run_queue_time"
 
 	MetricNameExtResContainerCpuTotalUsage MetricName = "ext_res_container_cpu_total_usage"
+	MetricNameExtCpuTotalDistribute        MetricName = "ext_cpu_total_distribute"
 )
 
 func GetCgroupPath(p *v1.Pod) string {