func (g *genericScheduler) Schedule(pod *api.Pod, minionLister algorithm.MinionLister) (string, error) {
minions, err := minionLister.List()
if err != nil {
return "", err
}
if len(minions.Items) == 0 {
return "", ErrNoNodesAvailable
}
filteredNodes, failedPredicateMap, err := findNodesThatFit(pod, g.pods, g.predicates, minions)
if err != nil {
return "", err
}
priorityList, err := PrioritizeNodes(pod, g.pods, g.prioritizers, algorithm.FakeMinionLister(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 minion or not, regardless of its value.
// If presence is true, prioritizes minions that have the specified label, regardless of value.
// If presence is false, prioritizes minions that do not have the specified label.
func (n *NodeLabelPrioritizer) CalculateNodeLabelPriority(pod *api.Pod, podLister algorithm.PodLister, minionLister algorithm.MinionLister) (algorithm.HostPriorityList, error) {
var score int
minions, err := minionLister.List()
if err != nil {
return nil, err
}
labeledMinions := map[string]bool{}
for _, minion := range minions.Items {
exists := labels.Set(minion.Labels).Has(n.label)
labeledMinions[minion.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 minionName, success := range labeledMinions {
if success {
score = 10
} else {
score = 0
}
result = append(result, algorithm.HostPriority{Host: minionName, Score: score})
}
return result, nil
}
// CalculateSpreadPriority spreads pods by minimizing the number of pods belonging to the same service
// on the same machine.
func (s *ServiceSpread) CalculateSpreadPriority(pod *api.Pod, podLister algorithm.PodLister, minionLister algorithm.MinionLister) (api.HostPriorityList, error) {
var maxCount int
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)
}
}
}
minions, err := minionLister.List()
if err != nil {
return nil, err
}
counts := map[string]int{}
if len(nsServicePods) > 0 {
for _, pod := range nsServicePods {
counts[pod.Spec.NodeName]++
// Compute the maximum number of pods hosted on any minion
if counts[pod.Spec.NodeName] > maxCount {
maxCount = counts[pod.Spec.NodeName]
}
}
}
result := []api.HostPriority{}
//score int - scale of 0-10
// 0 being the lowest priority and 10 being the highest
for _, minion := range minions.Items {
// initializing to the default/max minion score of 10
fScore := float32(10)
if maxCount > 0 {
fScore = 10 * (float32(maxCount-counts[minion.Name]) / float32(maxCount))
}
result = append(result, api.HostPriority{Host: minion.Name, Score: int(fScore)})
glog.V(10).Infof(
"%v -> %v: ServiceSpreadPriority, Score: (%d)", pod.Name, minion.Name, int(fScore),
)
}
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, minionLister algorithm.MinionLister) (algorithm.HostPriorityList, error) {
nodes, err := minionLister.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
}
// LeastRequestedPriority is a priority function that favors nodes with fewer requested resources.
// It calculates the percentage of memory and CPU requested by pods scheduled on the node, and prioritizes
// based on the minimum of the average of the fraction of requested to capacity.
// Details: cpu((capacity - sum(requested)) * 10 / capacity) + memory((capacity - sum(requested)) * 10 / capacity) / 2
func LeastRequestedPriority(pod *api.Pod, podLister algorithm.PodLister, minionLister algorithm.MinionLister) (api.HostPriorityList, error) {
nodes, err := minionLister.List()
if err != nil {
return api.HostPriorityList{}, err
}
podsToMachines, err := predicates.MapPodsToMachines(podLister)
list := api.HostPriorityList{}
for _, node := range nodes.Items {
list = append(list, calculateResourceOccupancy(pod, node, podsToMachines[node.Name]))
}
return list, nil
}
// Prioritizes the minions by running the individual priority functions sequentially.
// Each priority function is expected to set a score of 0-10
// 0 is the lowest priority score (least preferred minion) and 10 is the highest
// Each priority function can also have its own weight
// The minion scores returned by the priority function are multiplied by the weights to get weighted scores
// All scores are finally combined (added) to get the total weighted scores of all minions
func PrioritizeNodes(pod *api.Pod, podLister algorithm.PodLister, priorityConfigs []algorithm.PriorityConfig, minionLister algorithm.MinionLister, cloud cloudprovider.Interface) (api.HostPriorityList, error) {
result := api.HostPriorityList{}
// If no priority configs are provided, then the EqualPriority function is applied
// This is required to generate the priority list in the required format
if len(priorityConfigs) == 0 {
return EqualPriority(pod, podLister, minionLister)
}
combinedScores := map[string]int{}
for _, priorityConfig := range priorityConfigs {
weight := priorityConfig.Weight
// skip the priority function if the weight is specified as 0
if weight == 0 {
continue
}
priorityFunc := priorityConfig.Function
prioritizedList, err := priorityFunc(pod, podLister, minionLister)
if err != nil {
return api.HostPriorityList{}, err
}
for _, hostEntry := range prioritizedList {
combinedScores[hostEntry.Host] += hostEntry.Score * weight
}
}
if cloud != nil {
ext, supported := cloud.SchedulerExtension()
if supported {
nodes, err := minionLister.List()
if err != nil {
return api.HostPriorityList{}, err
}
prioritizedList, err := ext.Prioritize(pod, &nodes)
if err != nil {
return api.HostPriorityList{}, err
}
for _, hostEntry := range *prioritizedList {
combinedScores[hostEntry.Host] += hostEntry.Score
}
}
}
for host, score := range combinedScores {
glog.V(10).Infof("Host %s Score %d", host, score)
result = append(result, api.HostPriority{Host: host, Score: score})
}
return result, nil
}
// EqualPriority is a prioritizer function that gives an equal weight of one to all nodes
func EqualPriority(_ *api.Pod, podLister algorithm.PodLister, minionLister algorithm.MinionLister) (algorithm.HostPriorityList, error) {
nodes, err := minionLister.List()
if err != nil {
fmt.Errorf("failed to list nodes: %v", err)
return []algorithm.HostPriority{}, err
}
result := []algorithm.HostPriority{}
for _, minion := range nodes.Items {
result = append(result, algorithm.HostPriority{
Host: minion.Name,
Score: 1,
})
}
return result, nil
}
func numericPriority(pod *api.Pod, podLister algorithm.PodLister, minionLister algorithm.MinionLister) (algorithm.HostPriorityList, error) {
nodes, err := minionLister.List()
result := []algorithm.HostPriority{}
if err != nil {
return nil, fmt.Errorf("failed to list nodes: %v", err)
}
for _, minion := range nodes.Items {
score, err := strconv.Atoi(minion.Name)
if err != nil {
return nil, err
}
result = append(result, algorithm.HostPriority{
Host: minion.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 *SelectorSpread) CalculateSpreadPriority(pod *api.Pod, podLister algorithm.PodLister, minionLister algorithm.MinionLister) (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)
}
}
}
minions, err := minionLister.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 minion
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 _, minion := range minions.Items {
// initializing to the default/max minion score of 10
fScore := float32(10)
if maxCount > 0 {
fScore = 10 * (float32(maxCount-counts[minion.Name]) / float32(maxCount))
}
result = append(result, algorithm.HostPriority{Host: minion.Name, Score: int(fScore)})
glog.V(10).Infof(
"%v -> %v: SelectorSpreadPriority, Score: (%d)", pod.Name, minion.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, minionLister algorithm.MinionLister) (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)
}
}
}
minions, err := minionLister.List()
if err != nil {
return nil, err
}
// separate out the minions that have the label from the ones that don't
otherMinions := []string{}
labeledMinions := map[string]string{}
for _, minion := range minions.Items {
if labels.Set(minion.Labels).Has(s.label) {
label := labels.Set(minion.Labels).Get(s.label)
labeledMinions[minion.Name] = label
} else {
otherMinions = append(otherMinions, minion.Name)
}
}
podCounts := map[string]int{}
for _, pod := range nsServicePods {
label, exists := labeledMinions[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 minion := range labeledMinions {
// initializing to the default/max minion score of 10
fScore := float32(10)
if numServicePods > 0 {
fScore = 10 * (float32(numServicePods-podCounts[labeledMinions[minion]]) / float32(numServicePods))
}
result = append(result, algorithm.HostPriority{Host: minion, Score: int(fScore)})
}
// add the open minions with a score of 0
for _, minion := range otherMinions {
result = append(result, algorithm.HostPriority{Host: minion, Score: 0})
}
return result, nil
}
