196 lines
6.9 KiB
Go
196 lines
6.9 KiB
Go
/*
|
|
Copyright 2014 The Kubernetes Authors.
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at
|
|
|
|
http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and
|
|
limitations under the License.
|
|
*/
|
|
|
|
package scheduler
|
|
|
|
import (
|
|
"time"
|
|
|
|
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
|
|
"k8s.io/apimachinery/pkg/util/wait"
|
|
"k8s.io/kubernetes/pkg/api/v1"
|
|
"k8s.io/kubernetes/pkg/client/record"
|
|
"k8s.io/kubernetes/plugin/pkg/scheduler/algorithm"
|
|
schedulerapi "k8s.io/kubernetes/plugin/pkg/scheduler/api"
|
|
"k8s.io/kubernetes/plugin/pkg/scheduler/metrics"
|
|
"k8s.io/kubernetes/plugin/pkg/scheduler/schedulercache"
|
|
"k8s.io/kubernetes/plugin/pkg/scheduler/util"
|
|
|
|
clientset "k8s.io/kubernetes/pkg/client/clientset_generated/clientset"
|
|
|
|
"github.com/golang/glog"
|
|
"k8s.io/apimachinery/pkg/util/sets"
|
|
"k8s.io/kubernetes/pkg/client/cache"
|
|
)
|
|
|
|
// Binder knows how to write a binding.
|
|
type Binder interface {
|
|
Bind(binding *v1.Binding) error
|
|
}
|
|
|
|
type PodConditionUpdater interface {
|
|
Update(pod *v1.Pod, podCondition *v1.PodCondition) error
|
|
}
|
|
|
|
// Scheduler watches for new unscheduled pods. It attempts to find
|
|
// nodes that they fit on and writes bindings back to the api server.
|
|
type Scheduler struct {
|
|
config *Config
|
|
}
|
|
|
|
// These are the functions which need to be provided in order to build a Scheduler configuration.
|
|
// An implementation of this can be seen in factory.go.
|
|
type Configurator interface {
|
|
GetPriorityFunctionConfigs(priorityKeys sets.String) ([]algorithm.PriorityConfig, error)
|
|
GetPriorityMetadataProducer() (algorithm.MetadataProducer, error)
|
|
GetPredicateMetadataProducer() (algorithm.MetadataProducer, error)
|
|
GetPredicates(predicateKeys sets.String) (map[string]algorithm.FitPredicate, error)
|
|
GetHardPodAffinitySymmetricWeight() int
|
|
GetFailureDomains() []string
|
|
GetSchedulerName() string
|
|
MakeDefaultErrorFunc(backoff *util.PodBackoff, podQueue *cache.FIFO) func(pod *v1.Pod, err error)
|
|
|
|
// Probably doesn't need to be public. But exposed for now in case.
|
|
ResponsibleForPod(pod *v1.Pod) bool
|
|
|
|
// Needs to be exposed for things like integration tests where we want to make fake nodes.
|
|
GetNodeStore() cache.Store
|
|
GetClient() clientset.Interface
|
|
GetScheduledPodListerIndexer() cache.Indexer
|
|
Run()
|
|
|
|
Create() (*Config, error)
|
|
CreateFromProvider(providerName string) (*Config, error)
|
|
CreateFromConfig(policy schedulerapi.Policy) (*Config, error)
|
|
CreateFromKeys(predicateKeys, priorityKeys sets.String, extenders []algorithm.SchedulerExtender) (*Config, error)
|
|
}
|
|
|
|
type Config struct {
|
|
// It is expected that changes made via SchedulerCache will be observed
|
|
// by NodeLister and Algorithm.
|
|
SchedulerCache schedulercache.Cache
|
|
NodeLister algorithm.NodeLister
|
|
Algorithm algorithm.ScheduleAlgorithm
|
|
Binder Binder
|
|
// PodConditionUpdater is used only in case of scheduling errors. If we succeed
|
|
// with scheduling, PodScheduled condition will be updated in apiserver in /bind
|
|
// handler so that binding and setting PodCondition it is atomic.
|
|
PodConditionUpdater PodConditionUpdater
|
|
|
|
// 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() *v1.Pod
|
|
|
|
// Error is called if there is an error. It is passed the pod in
|
|
// question, and the error
|
|
Error func(*v1.Pod, error)
|
|
|
|
// Recorder is the EventRecorder to use
|
|
Recorder record.EventRecorder
|
|
|
|
// Close this to shut down the scheduler.
|
|
StopEverything chan struct{}
|
|
}
|
|
|
|
// New returns a new scheduler.
|
|
func New(c *Config) *Scheduler {
|
|
s := &Scheduler{
|
|
config: c,
|
|
}
|
|
metrics.Register()
|
|
return s
|
|
}
|
|
|
|
// Run begins watching and scheduling. It starts a goroutine and returns immediately.
|
|
func (s *Scheduler) Run() {
|
|
go wait.Until(s.scheduleOne, 0, s.config.StopEverything)
|
|
}
|
|
|
|
func (s *Scheduler) scheduleOne() {
|
|
pod := s.config.NextPod()
|
|
|
|
glog.V(3).Infof("Attempting to schedule pod: %v/%v", pod.Namespace, pod.Name)
|
|
start := time.Now()
|
|
dest, err := s.config.Algorithm.Schedule(pod, s.config.NodeLister)
|
|
if err != nil {
|
|
glog.V(1).Infof("Failed to schedule pod: %v/%v", pod.Namespace, pod.Name)
|
|
s.config.Error(pod, err)
|
|
s.config.Recorder.Eventf(pod, v1.EventTypeWarning, "FailedScheduling", "%v", err)
|
|
s.config.PodConditionUpdater.Update(pod, &v1.PodCondition{
|
|
Type: v1.PodScheduled,
|
|
Status: v1.ConditionFalse,
|
|
Reason: v1.PodReasonUnschedulable,
|
|
Message: err.Error(),
|
|
})
|
|
return
|
|
}
|
|
metrics.SchedulingAlgorithmLatency.Observe(metrics.SinceInMicroseconds(start))
|
|
|
|
// Optimistically assume that the binding will succeed and send it to apiserver
|
|
// in the background.
|
|
// If the binding fails, scheduler will release resources allocated to assumed pod
|
|
// immediately.
|
|
assumed := *pod
|
|
assumed.Spec.NodeName = dest
|
|
if err := s.config.SchedulerCache.AssumePod(&assumed); err != nil {
|
|
glog.Errorf("scheduler cache AssumePod failed: %v", err)
|
|
// TODO: This means that a given pod is already in cache (which means it
|
|
// is either assumed or already added). This is most probably result of a
|
|
// BUG in retrying logic. As a temporary workaround (which doesn't fully
|
|
// fix the problem, but should reduce its impact), we simply return here,
|
|
// as binding doesn't make sense anyway.
|
|
// This should be fixed properly though.
|
|
return
|
|
}
|
|
|
|
go func() {
|
|
defer metrics.E2eSchedulingLatency.Observe(metrics.SinceInMicroseconds(start))
|
|
|
|
b := &v1.Binding{
|
|
ObjectMeta: metav1.ObjectMeta{Namespace: pod.Namespace, Name: pod.Name},
|
|
Target: v1.ObjectReference{
|
|
Kind: "Node",
|
|
Name: dest,
|
|
},
|
|
}
|
|
|
|
bindingStart := time.Now()
|
|
// If binding succeeded then PodScheduled condition will be updated in apiserver so that
|
|
// it's atomic with setting host.
|
|
err := s.config.Binder.Bind(b)
|
|
if err := s.config.SchedulerCache.FinishBinding(&assumed); err != nil {
|
|
glog.Errorf("scheduler cache FinishBinding failed: %v", err)
|
|
}
|
|
if err != nil {
|
|
glog.V(1).Infof("Failed to bind pod: %v/%v", pod.Namespace, pod.Name)
|
|
if err := s.config.SchedulerCache.ForgetPod(&assumed); err != nil {
|
|
glog.Errorf("scheduler cache ForgetPod failed: %v", err)
|
|
}
|
|
s.config.Error(pod, err)
|
|
s.config.Recorder.Eventf(pod, v1.EventTypeNormal, "FailedScheduling", "Binding rejected: %v", err)
|
|
s.config.PodConditionUpdater.Update(pod, &v1.PodCondition{
|
|
Type: v1.PodScheduled,
|
|
Status: v1.ConditionFalse,
|
|
Reason: "BindingRejected",
|
|
})
|
|
return
|
|
}
|
|
metrics.BindingLatency.Observe(metrics.SinceInMicroseconds(bindingStart))
|
|
s.config.Recorder.Eventf(pod, v1.EventTypeNormal, "Scheduled", "Successfully assigned %v to %v", pod.Name, dest)
|
|
}()
|
|
}
|