Documentation ¶
Index ¶
- Variables
- func EqualPriority(_ *api.Pod, podLister algorithm.PodLister, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error)
- func NewGenericScheduler(predicates map[string]algorithm.FitPredicate, ...) algorithm.ScheduleAlgorithm
- func PrioritizeNodes(pod *api.Pod, podLister algorithm.PodLister, ...) (algorithm.HostPriorityList, error)
- type Binder
- type Config
- type ExtendedPodLister
- type FailedPredicateMap
- type FakeModeler
- type FitError
- type Scheduler
- type SimpleModeler
- type SystemModeler
Constants ¶
This section is empty.
Variables ¶
var ErrNoNodesAvailable = fmt.Errorf("no nodes available to schedule pods")
Functions ¶
func EqualPriority ¶ added in v0.21.2
func EqualPriority(_ *api.Pod, podLister algorithm.PodLister, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error)
EqualPriority is a prioritizer function that gives an equal weight of one to all nodes
func NewGenericScheduler ¶ added in v0.21.2
func NewGenericScheduler(predicates map[string]algorithm.FitPredicate, prioritizers []algorithm.PriorityConfig, pods algorithm.PodLister, random *rand.Rand) algorithm.ScheduleAlgorithm
func PrioritizeNodes ¶ added in v0.21.2
func PrioritizeNodes(pod *api.Pod, podLister algorithm.PodLister, priorityConfigs []algorithm.PriorityConfig, nodeLister algorithm.NodeLister) (algorithm.HostPriorityList, error)
Prioritizes the nodes 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 node) and 10 is the highest Each priority function can also have its own weight The node 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 nodes
Types ¶
type Config ¶
type Config struct { // It is expected that changes made via modeler will be observed // by NodeLister and Algorithm. Modeler SystemModeler NodeLister algorithm.NodeLister Algorithm algorithm.ScheduleAlgorithm Binder Binder // Rate at which we can create pods BindPodsRateLimiter util.RateLimiter // NextPod should be a function that blocks until the next pod // is available. We don't use a channel for this, because scheduling // a pod may take some amount of time and we don't want pods to get // stale while they sit in a channel. NextPod func() *api.Pod // Error is called if there is an error. It is passed the pod in // question, and the error Error func(*api.Pod, error) // Recorder is the EventRecorder to use Recorder record.EventRecorder // Close this to shut down the scheduler. StopEverything chan struct{} }
type ExtendedPodLister ¶
ExtendedPodLister: SimpleModeler needs to be able to check for a pod's existence in addition to listing the pods.
type FailedPredicateMap ¶ added in v0.21.2
type FakeModeler ¶
type FakeModeler struct { AssumePodFunc func(pod *api.Pod) ForgetPodFunc func(pod *api.Pod) ForgetPodByKeyFunc func(key string) // contains filtered or unexported fields }
FakeModeler implements the SystemModeler interface.
func (*FakeModeler) AssumePod ¶
func (f *FakeModeler) AssumePod(pod *api.Pod)
AssumePod calls the function variable if it is not nil.
func (*FakeModeler) ForgetPod ¶
func (f *FakeModeler) ForgetPod(pod *api.Pod)
ForgetPod calls the function variable if it is not nil.
func (*FakeModeler) ForgetPodByKey ¶
func (f *FakeModeler) ForgetPodByKey(key string)
ForgetPodByKey calls the function variable if it is not nil.
func (*FakeModeler) LockedAction ¶
func (a *FakeModeler) LockedAction(do func())
LockedAction serializes calls of whatever is passed as 'do'.
type FitError ¶ added in v0.21.2
type FitError struct { Pod *api.Pod FailedPredicates FailedPredicateMap }
type Scheduler ¶
type Scheduler struct {
// contains filtered or unexported fields
}
Scheduler watches for new unscheduled pods. It attempts to find nodes that they fit on and writes bindings back to the api server.
type SimpleModeler ¶
type SimpleModeler struct {
// contains filtered or unexported fields
}
SimpleModeler implements the SystemModeler interface with a timed pod cache.
func NewSimpleModeler ¶
func NewSimpleModeler(queuedPods, scheduledPods ExtendedPodLister) *SimpleModeler
NewSimpleModeler returns a new SimpleModeler.
queuedPods: a PodLister that will return pods that have not been scheduled yet. scheduledPods: a PodLister that will return pods that we know for sure have been scheduled.
func (*SimpleModeler) AssumePod ¶
func (s *SimpleModeler) AssumePod(pod *api.Pod)
func (*SimpleModeler) ForgetPod ¶
func (s *SimpleModeler) ForgetPod(pod *api.Pod)
func (*SimpleModeler) ForgetPodByKey ¶
func (s *SimpleModeler) ForgetPodByKey(key string)
func (*SimpleModeler) LockedAction ¶
func (a *SimpleModeler) LockedAction(do func())
LockedAction serializes calls of whatever is passed as 'do'.
func (*SimpleModeler) PodLister ¶
func (s *SimpleModeler) PodLister() algorithm.PodLister
PodLister returns a PodLister that will list pods that we think we have scheduled in addition to pods that we know have been scheduled.
type SystemModeler ¶
type SystemModeler interface { // AssumePod assumes that the given pod exists in the system. // The assumtion should last until the system confirms the // assumtion or disconfirms it. AssumePod(pod *api.Pod) // ForgetPod removes a pod assumtion. (It won't make the model // show the absence of the given pod if the pod is in the scheduled // pods list!) ForgetPod(pod *api.Pod) ForgetPodByKey(key string) // For serializing calls to Assume/ForgetPod: imagine you want to add // a pod if and only if a bind succeeds, but also remove a pod if it is deleted. // TODO: if SystemModeler begins modeling things other than pods, this // should probably be parameterized or specialized for pods. LockedAction(f func()) }
SystemModeler can help scheduler produce a model of the system that anticipates reality. For example, if scheduler has pods A and B both using hostPort 80, when it binds A to machine M it should not bind B to machine M in the time when it hasn't observed the binding of A take effect yet.
Since the model is only an optimization, it's expected to handle any errors itself without sending them back to the scheduler.
Directories ¶
Path | Synopsis |
---|---|
Package scheduler contains a generic Scheduler interface and several implementations.
|
Package scheduler contains a generic Scheduler interface and several implementations. |
This package is used to register algorithm provider plugins.
|
This package is used to register algorithm provider plugins. |
defaults
This is the default algorithm provider for the scheduler.
|
This is the default algorithm provider for the scheduler. |
Package factory can set up a scheduler.
|
Package factory can set up a scheduler. |