func (g *genericScheduler) Schedule(pod *api.Pod, nodeLister algorithm.NodeLister) (string, error) {
nodes, err := nodeLister.List()
if err != nil {
return "", err
}
if len(nodes.Items) == 0 {
return "", ErrNoNodesAvailable
}
filteredNodes, failedPredicateMap, err := findNodesThatFit(pod, g.pods, g.predicates, nodes)
if err != nil {
return "", err
}
priorityList, err := PrioritizeNodes(pod, g.pods, g.prioritizers, algorithm.FakeNodeLister(filteredNodes))
if err != nil {
return "", err
}
if len(priorityList) == 0 {
return "", &FitError{
Pod: pod,
FailedPredicates: failedPredicateMap,
}
}
return g.selectHost(priorityList)
}
// CalculateNodeLabelPriority checks whether a particular label exists on a node or not, regardless of its value.
// If presence is true, prioritizes nodes that have the specified label, regardless of value.
// If presence is false, prioritizes nodes that do not have the specified label.
func (n *NodeLabelPrioritizer) CalculateNodeLabelPriority(pod *api.Pod, podLister algorithm.PodLister, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error) {
var score int
nodes, err := nodeLister.List()
if err != nil {
return nil, err
}
labeledNodes := map[string]bool{}
for _, node := range nodes.Items {
exists := labels.Set(node.Labels).Has(n.label)
labeledNodes[node.Name] = (exists && n.presence) || (!exists && !n.presence)
}
result := []algorithm.HostPriority{}
//score int - scale of 0-10
// 0 being the lowest priority and 10 being the highest
for nodeName, success := range labeledNodes {
if success {
score = 10
} else {
score = 0
}
result = append(result, algorithm.HostPriority{Host: nodeName, Score: score})
}
return result, nil
}
// BalancedResourceAllocation favors nodes with balanced resource usage rate.
// BalancedResourceAllocation should **NOT** be used alone, and **MUST** be used together with LeastRequestedPriority.
// It calculates the difference between the cpu and memory fracion of capacity, and prioritizes the host based on how
// close the two metrics are to each other.
// Detail: score = 10 - abs(cpuFraction-memoryFraction)*10. The algorithm is partly inspired by:
// "Wei Huang et al. An Energy Efficient Virtual Machine Placement Algorithm with Balanced Resource Utilization"
func BalancedResourceAllocation(pod *api.Pod, podLister algorithm.PodLister, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error) {
nodes, err := nodeLister.List()
if err != nil {
return algorithm.HostPriorityList{}, err
}
podsToMachines, err := predicates.MapPodsToMachines(podLister)
list := algorithm.HostPriorityList{}
for _, node := range nodes.Items {
list = append(list, calculateBalancedResourceAllocation(pod, node, podsToMachines[node.Name]))
}
return list, nil
}
// EqualPriority is a prioritizer function that gives an equal weight of one to all nodes
func EqualPriority(_ *api.Pod, podLister algorithm.PodLister, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error) {
nodes, err := nodeLister.List()
if err != nil {
glog.Errorf("failed to list nodes: %v", err)
return []algorithm.HostPriority{}, err
}
result := []algorithm.HostPriority{}
for _, node := range nodes.Items {
result = append(result, algorithm.HostPriority{
Host: node.Name,
Score: 1,
})
}
return result, nil
}
func numericPriority(pod *api.Pod, podLister algorithm.PodLister, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error) {
nodes, err := nodeLister.List()
result := []algorithm.HostPriority{}
if err != nil {
return nil, fmt.Errorf("failed to list nodes: %v", err)
}
for _, node := range nodes.Items {
score, err := strconv.Atoi(node.Name)
if err != nil {
return nil, err
}
result = append(result, algorithm.HostPriority{
Host: node.Name,
Score: score,
})
}
return result, nil
}
// CalculateSpreadPriority spreads pods by minimizing the number of pods belonging to the same service or replication controller. It counts number of pods that run under
// Services or RCs as the pod being scheduled and tries to minimize the number of conflicts. I.e. pushes scheduler towards a Node where there's a smallest number of
// pods which match the same selectors of Services and RCs as current pod.
func (s *NetworkSpread) CalculateSpreadPriority(pod *api.Pod, podLister algorithm.PodLister, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error) {
var maxCount int
var nsPods []*api.Pod
selectors := make([]labels.Selector, 0)
services, err := s.serviceLister.GetPodServices(pod)
if err == nil {
for _, service := range services {
selectors = append(selectors, labels.SelectorFromSet(service.Spec.Selector))
}
}
controllers, err := s.controllerLister.GetPodControllers(pod)
if err == nil {
for _, controller := range controllers {
selectors = append(selectors, labels.SelectorFromSet(controller.Spec.Selector))
}
}
if len(selectors) > 0 {
pods, err := podLister.List(labels.Everything())
if err != nil {
return nil, err
}
// consider only the pods that belong to the same namespace
for _, nsPod := range pods {
if nsPod.Namespace == pod.Namespace {
nsPods = append(nsPods, nsPod)
}
}
}
nodes, err := nodeLister.List()
if err != nil {
return nil, err
}
subnetMap := map[string]string{}
for _, node := range nodes.Items {
for _, addrs := range node.Status.Addresses {
if addrs.Type == api.NodeLegacyHostIP {
// for now wandoujia used 24 mask, a pit here.
_, subnet, _ := net.ParseCIDR(fmt.Sprintf("%s/24", addrs.Address))
subnetMap[node.Name] = subnet.String()
}
}
}
counts := map[string]int{}
if len(nsPods) > 0 {
for _, pod := range nsPods {
matches := false
for _, selector := range selectors {
if selector.Matches(labels.Set(pod.ObjectMeta.Labels)) {
matches = true
break
}
}
if matches {
subnet := subnetMap[pod.Spec.NodeName]
counts[subnet]++
// Compute the maximum number of pods hosted on any node
if counts[subnet] > maxCount {
maxCount = counts[subnet]
}
}
}
}
result := []algorithm.HostPriority{}
//score int - scale of 0-10
// 0 being the lowest priority and 10 being the highest
for _, node := range nodes.Items {
// initializing to the default/max node score of 10
fScore := float32(10)
subnet := subnetMap[node.Name]
if maxCount > 0 {
fScore = 10 * (float32(maxCount-counts[subnet]) / float32(maxCount))
}
glog.Infof("maxcount: %v, subNetCounts: %v, fscore: %v", maxCount, counts[subnet], fScore)
result = append(result, algorithm.HostPriority{Host: node.Name, Score: int(fScore)})
glog.V(10).Infof(
"%v -> %v: SelectorSpreadPriority, Score: (%d)", pod.Name, node.Name, int(fScore),
)
}
return result, nil
}
// CalculateSpreadPriority spreads pods by minimizing the number of pods belonging to the same service or replication controller. It counts number of pods that run under
// Services or RCs as the pod being scheduled and tries to minimize the number of conflicts. I.e. pushes scheduler towards a Node where there's a smallest number of
// pods which match the same selectors of Services and RCs as current pod.
func (s *SelectorSpread) CalculateSpreadPriority(pod *api.Pod, podLister algorithm.PodLister, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error) {
var maxCount int
var nsPods []*api.Pod
selectors := make([]labels.Selector, 0)
services, err := s.serviceLister.GetPodServices(pod)
if err == nil {
for _, service := range services {
selectors = append(selectors, labels.SelectorFromSet(service.Spec.Selector))
}
}
controllers, err := s.controllerLister.GetPodControllers(pod)
if err == nil {
for _, controller := range controllers {
selectors = append(selectors, labels.SelectorFromSet(controller.Spec.Selector))
}
}
if len(selectors) > 0 {
pods, err := podLister.List(labels.Everything())
if err != nil {
return nil, err
}
// consider only the pods that belong to the same namespace
for _, nsPod := range pods {
if nsPod.Namespace == pod.Namespace {
nsPods = append(nsPods, nsPod)
}
}
}
nodes, err := nodeLister.List()
if err != nil {
return nil, err
}
counts := map[string]int{}
if len(nsPods) > 0 {
for _, pod := range nsPods {
matches := false
for _, selector := range selectors {
if selector.Matches(labels.Set(pod.ObjectMeta.Labels)) {
matches = true
break
}
}
if matches {
counts[pod.Spec.NodeName]++
// Compute the maximum number of pods hosted on any node
if counts[pod.Spec.NodeName] > maxCount {
maxCount = counts[pod.Spec.NodeName]
}
}
}
}
result := []algorithm.HostPriority{}
//score int - scale of 0-10
// 0 being the lowest priority and 10 being the highest
for _, node := range nodes.Items {
// initializing to the default/max node score of 10
fScore := float32(10)
if maxCount > 0 {
fScore = 10 * (float32(maxCount-counts[node.Name]) / float32(maxCount))
}
result = append(result, algorithm.HostPriority{Host: node.Name, Score: int(fScore)})
glog.V(10).Infof(
"%v -> %v: SelectorSpreadPriority, Score: (%d)", pod.Name, node.Name, int(fScore),
)
}
return result, nil
}
// CalculateAntiAffinityPriority spreads pods by minimizing the number of pods belonging to the same service
// on machines with the same value for a particular label.
// The label to be considered is provided to the struct (ServiceAntiAffinity).
func (s *ServiceAntiAffinity) CalculateAntiAffinityPriority(pod *api.Pod, podLister algorithm.PodLister, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error) {
var nsServicePods []*api.Pod
services, err := s.serviceLister.GetPodServices(pod)
if err == nil {
// just use the first service and get the other pods within the service
// TODO: a separate predicate can be created that tries to handle all services for the pod
selector := labels.SelectorFromSet(services[0].Spec.Selector)
pods, err := podLister.List(selector)
if err != nil {
return nil, err
}
// consider only the pods that belong to the same namespace
for _, nsPod := range pods {
if nsPod.Namespace == pod.Namespace {
nsServicePods = append(nsServicePods, nsPod)
}
}
}
nodes, err := nodeLister.List()
if err != nil {
return nil, err
}
// separate out the nodes that have the label from the ones that don't
otherNodes := []string{}
labeledNodes := map[string]string{}
for _, node := range nodes.Items {
if labels.Set(node.Labels).Has(s.label) {
label := labels.Set(node.Labels).Get(s.label)
labeledNodes[node.Name] = label
} else {
otherNodes = append(otherNodes, node.Name)
}
}
podCounts := map[string]int{}
for _, pod := range nsServicePods {
label, exists := labeledNodes[pod.Spec.NodeName]
if !exists {
continue
}
podCounts[label]++
}
numServicePods := len(nsServicePods)
result := []algorithm.HostPriority{}
//score int - scale of 0-10
// 0 being the lowest priority and 10 being the highest
for node := range labeledNodes {
// initializing to the default/max node score of 10
fScore := float32(10)
if numServicePods > 0 {
fScore = 10 * (float32(numServicePods-podCounts[labeledNodes[node]]) / float32(numServicePods))
}
result = append(result, algorithm.HostPriority{Host: node, Score: int(fScore)})
}
// add the open nodes with a score of 0
for _, node := range otherNodes {
result = append(result, algorithm.HostPriority{Host: node, Score: 0})
}
return result, nil
}