Fresh dep ensure

This commit is contained in:
Mike Cronce
2018-11-26 13:23:56 -05:00
parent 93cb8a04d7
commit 407478ab9a
9016 changed files with 551394 additions and 279685 deletions

View File

@ -1,36 +1,4 @@
package(default_visibility = ["//visibility:public"])
load(
"@io_bazel_rules_go//go:def.bzl",
"go_library",
"go_test",
)
go_test(
name = "go_default_test",
srcs = ["scheduler_test.go"],
embed = [":go_default_library"],
deps = [
"//pkg/api/legacyscheme:go_default_library",
"//pkg/controller/volume/persistentvolume:go_default_library",
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/predicates:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/core:go_default_library",
"//pkg/scheduler/testing:go_default_library",
"//pkg/scheduler/volumebinder:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/types:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/diff:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/wait:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//vendor/k8s.io/client-go/tools/cache:go_default_library",
"//vendor/k8s.io/client-go/tools/record:go_default_library",
],
)
load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
go_library(
name = "go_default_library",
@ -39,25 +7,74 @@ go_library(
"testutil.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler",
visibility = ["//visibility:public"],
deps = [
"//pkg/features:go_default_library",
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/predicates:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/api/latest:go_default_library",
"//pkg/scheduler/apis/config:go_default_library",
"//pkg/scheduler/core:go_default_library",
"//pkg/scheduler/factory:go_default_library",
"//pkg/scheduler/internal/cache:go_default_library",
"//pkg/scheduler/internal/queue:go_default_library",
"//pkg/scheduler/metrics:go_default_library",
"//pkg/scheduler/util:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/wait:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/client-go/informers/apps/v1:go_default_library",
"//staging/src/k8s.io/client-go/informers/core/v1:go_default_library",
"//staging/src/k8s.io/client-go/informers/policy/v1beta1:go_default_library",
"//staging/src/k8s.io/client-go/informers/storage/v1:go_default_library",
"//staging/src/k8s.io/client-go/kubernetes:go_default_library",
"//staging/src/k8s.io/client-go/listers/core/v1:go_default_library",
"//staging/src/k8s.io/client-go/tools/record:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = [
"main_test.go",
"scheduler_test.go",
],
embed = [":go_default_library"],
deps = [
"//pkg/api/legacyscheme:go_default_library",
"//pkg/controller/volume/persistentvolume:go_default_library",
"//pkg/features:go_default_library",
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/predicates:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/apis/config:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/core:go_default_library",
"//pkg/scheduler/factory:go_default_library",
"//pkg/scheduler/internal/cache:go_default_library",
"//pkg/scheduler/internal/cache/fake:go_default_library",
"//pkg/scheduler/volumebinder:go_default_library",
"//vendor/github.com/golang/glog:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/wait:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//vendor/k8s.io/client-go/kubernetes:go_default_library",
"//vendor/k8s.io/client-go/listers/core/v1:go_default_library",
"//vendor/k8s.io/client-go/tools/record:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/types:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/diff:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/wait:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature/testing:go_default_library",
"//staging/src/k8s.io/client-go/informers:go_default_library",
"//staging/src/k8s.io/client-go/kubernetes/fake:go_default_library",
"//staging/src/k8s.io/client-go/listers/core/v1:go_default_library",
"//staging/src/k8s.io/client-go/tools/cache:go_default_library",
"//staging/src/k8s.io/client-go/tools/record:go_default_library",
],
)
@ -75,13 +92,17 @@ filegroup(
"//pkg/scheduler/algorithm:all-srcs",
"//pkg/scheduler/algorithmprovider:all-srcs",
"//pkg/scheduler/api:all-srcs",
"//pkg/scheduler/apis/config:all-srcs",
"//pkg/scheduler/cache:all-srcs",
"//pkg/scheduler/core:all-srcs",
"//pkg/scheduler/factory:all-srcs",
"//pkg/scheduler/internal/cache:all-srcs",
"//pkg/scheduler/internal/queue:all-srcs",
"//pkg/scheduler/metrics:all-srcs",
"//pkg/scheduler/testing:all-srcs",
"//pkg/scheduler/util:all-srcs",
"//pkg/scheduler/volumebinder:all-srcs",
],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

View File

@ -12,31 +12,27 @@ go_library(
"doc.go",
"scheduler_interface.go",
"types.go",
"well_known_labels.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/algorithm",
deps = [
"//pkg/apis/core:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//vendor/k8s.io/api/apps/v1beta1:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/extensions/v1beta1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//pkg/scheduler/internal/cache:go_default_library",
"//staging/src/k8s.io/api/apps/v1:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/api/policy/v1beta1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = [
"scheduler_interface_test.go",
"types_test.go",
],
srcs = ["types_test.go"],
embed = [":go_default_library"],
deps = [
"//pkg/scheduler/cache:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
],
)

View File

@ -9,6 +9,7 @@ load(
go_library(
name = "go_default_library",
srcs = [
"csi_volume_predicate.go",
"error.go",
"metadata.go",
"predicates.go",
@ -23,29 +24,32 @@ go_library(
"//pkg/kubelet/apis:go_default_library",
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/priorities/util:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/util:go_default_library",
"//pkg/scheduler/volumebinder:go_default_library",
"//pkg/volume/util:go_default_library",
"//vendor/github.com/golang/glog:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/storage/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/errors:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/fields:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/rand:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//vendor/k8s.io/client-go/listers/core/v1:go_default_library",
"//vendor/k8s.io/client-go/listers/storage/v1:go_default_library",
"//vendor/k8s.io/client-go/util/workqueue:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/api/storage/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/errors:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/fields:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/rand:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/client-go/listers/core/v1:go_default_library",
"//staging/src/k8s.io/client-go/listers/storage/v1:go_default_library",
"//staging/src/k8s.io/client-go/util/workqueue:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = [
"csi_volume_predicate_test.go",
"main_test.go",
"max_attachable_volume_predicate_test.go",
"metadata_test.go",
"predicates_test.go",
@ -57,17 +61,18 @@ go_test(
"//pkg/features:go_default_library",
"//pkg/kubelet/apis:go_default_library",
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/testing:go_default_library",
"//pkg/volume/util:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/storage/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature/testing:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/api/storage/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature/testing:go_default_library",
],
)

View File

@ -0,0 +1,157 @@
/*
Copyright 2018 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 predicates
import (
"fmt"
"k8s.io/api/core/v1"
utilfeature "k8s.io/apiserver/pkg/util/feature"
"k8s.io/klog"
"k8s.io/kubernetes/pkg/features"
"k8s.io/kubernetes/pkg/scheduler/algorithm"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
volumeutil "k8s.io/kubernetes/pkg/volume/util"
)
// CSIMaxVolumeLimitChecker defines predicate needed for counting CSI volumes
type CSIMaxVolumeLimitChecker struct {
pvInfo PersistentVolumeInfo
pvcInfo PersistentVolumeClaimInfo
}
// NewCSIMaxVolumeLimitPredicate returns a predicate for counting CSI volumes
func NewCSIMaxVolumeLimitPredicate(
pvInfo PersistentVolumeInfo, pvcInfo PersistentVolumeClaimInfo) algorithm.FitPredicate {
c := &CSIMaxVolumeLimitChecker{
pvInfo: pvInfo,
pvcInfo: pvcInfo,
}
return c.attachableLimitPredicate
}
func (c *CSIMaxVolumeLimitChecker) attachableLimitPredicate(
pod *v1.Pod, meta algorithm.PredicateMetadata, nodeInfo *schedulercache.NodeInfo) (bool, []algorithm.PredicateFailureReason, error) {
// if feature gate is disable we return
if !utilfeature.DefaultFeatureGate.Enabled(features.AttachVolumeLimit) {
return true, nil, nil
}
// If a pod doesn't have any volume attached to it, the predicate will always be true.
// Thus we make a fast path for it, to avoid unnecessary computations in this case.
if len(pod.Spec.Volumes) == 0 {
return true, nil, nil
}
nodeVolumeLimits := nodeInfo.VolumeLimits()
// if node does not have volume limits this predicate should exit
if len(nodeVolumeLimits) == 0 {
return true, nil, nil
}
// a map of unique volume name/csi volume handle and volume limit key
newVolumes := make(map[string]string)
if err := c.filterAttachableVolumes(pod.Spec.Volumes, pod.Namespace, newVolumes); err != nil {
return false, nil, err
}
if len(newVolumes) == 0 {
return true, nil, nil
}
// a map of unique volume name/csi volume handle and volume limit key
attachedVolumes := make(map[string]string)
for _, existingPod := range nodeInfo.Pods() {
if err := c.filterAttachableVolumes(existingPod.Spec.Volumes, existingPod.Namespace, attachedVolumes); err != nil {
return false, nil, err
}
}
newVolumeCount := map[string]int{}
attachedVolumeCount := map[string]int{}
for volumeName, volumeLimitKey := range attachedVolumes {
if _, ok := newVolumes[volumeName]; ok {
delete(newVolumes, volumeName)
}
attachedVolumeCount[volumeLimitKey]++
}
for _, volumeLimitKey := range newVolumes {
newVolumeCount[volumeLimitKey]++
}
for volumeLimitKey, count := range newVolumeCount {
maxVolumeLimit, ok := nodeVolumeLimits[v1.ResourceName(volumeLimitKey)]
if ok {
currentVolumeCount := attachedVolumeCount[volumeLimitKey]
if currentVolumeCount+count > int(maxVolumeLimit) {
return false, []algorithm.PredicateFailureReason{ErrMaxVolumeCountExceeded}, nil
}
}
}
return true, nil, nil
}
func (c *CSIMaxVolumeLimitChecker) filterAttachableVolumes(
volumes []v1.Volume, namespace string, result map[string]string) error {
for _, vol := range volumes {
// CSI volumes can only be used as persistent volumes
if vol.PersistentVolumeClaim == nil {
continue
}
pvcName := vol.PersistentVolumeClaim.ClaimName
if pvcName == "" {
return fmt.Errorf("PersistentVolumeClaim had no name")
}
pvc, err := c.pvcInfo.GetPersistentVolumeClaimInfo(namespace, pvcName)
if err != nil {
klog.V(4).Infof("Unable to look up PVC info for %s/%s", namespace, pvcName)
continue
}
pvName := pvc.Spec.VolumeName
// TODO - the actual handling of unbound PVCs will be fixed by late binding design.
if pvName == "" {
klog.V(4).Infof("Persistent volume had no name for claim %s/%s", namespace, pvcName)
continue
}
pv, err := c.pvInfo.GetPersistentVolumeInfo(pvName)
if err != nil {
klog.V(4).Infof("Unable to look up PV info for PVC %s/%s and PV %s", namespace, pvcName, pvName)
continue
}
csiSource := pv.Spec.PersistentVolumeSource.CSI
if csiSource == nil {
klog.V(4).Infof("Not considering non-CSI volume %s/%s", namespace, pvcName)
continue
}
driverName := csiSource.Driver
volumeLimitKey := volumeutil.GetCSIAttachLimitKey(driverName)
result[csiSource.VolumeHandle] = volumeLimitKey
}
return nil
}

View File

@ -0,0 +1,179 @@
/*
Copyright 2018 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 predicates
import (
"reflect"
"testing"
"k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
utilfeature "k8s.io/apiserver/pkg/util/feature"
utilfeaturetesting "k8s.io/apiserver/pkg/util/feature/testing"
"k8s.io/kubernetes/pkg/features"
"k8s.io/kubernetes/pkg/scheduler/algorithm"
)
func TestCSIVolumeCountPredicate(t *testing.T) {
// for pods with CSI pvcs
oneVolPod := &v1.Pod{
Spec: v1.PodSpec{
Volumes: []v1.Volume{
{
VolumeSource: v1.VolumeSource{
PersistentVolumeClaim: &v1.PersistentVolumeClaimVolumeSource{
ClaimName: "csi-ebs",
},
},
},
},
},
}
twoVolPod := &v1.Pod{
Spec: v1.PodSpec{
Volumes: []v1.Volume{
{
VolumeSource: v1.VolumeSource{
PersistentVolumeClaim: &v1.PersistentVolumeClaimVolumeSource{
ClaimName: "cs-ebs-1",
},
},
},
{
VolumeSource: v1.VolumeSource{
PersistentVolumeClaim: &v1.PersistentVolumeClaimVolumeSource{
ClaimName: "csi-ebs-2",
},
},
},
},
},
}
runningPod := &v1.Pod{
Spec: v1.PodSpec{
Volumes: []v1.Volume{
{
VolumeSource: v1.VolumeSource{
PersistentVolumeClaim: &v1.PersistentVolumeClaimVolumeSource{
ClaimName: "csi-ebs-3",
},
},
},
},
},
}
tests := []struct {
newPod *v1.Pod
existingPods []*v1.Pod
filterName string
maxVols int
fits bool
test string
}{
{
newPod: oneVolPod,
existingPods: []*v1.Pod{runningPod, twoVolPod},
filterName: "csi-ebs",
maxVols: 4,
fits: true,
test: "fits when node capacity >= new pods CSI volume",
},
{
newPod: oneVolPod,
existingPods: []*v1.Pod{runningPod, twoVolPod},
filterName: "csi-ebs",
maxVols: 2,
fits: false,
test: "doesn't when node capacity <= pods CSI volume",
},
}
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.AttachVolumeLimit, true)()
expectedFailureReasons := []algorithm.PredicateFailureReason{ErrMaxVolumeCountExceeded}
// running attachable predicate tests with feature gate and limit present on nodes
for _, test := range tests {
node := getNodeWithPodAndVolumeLimits(test.existingPods, int64(test.maxVols), test.filterName)
pred := NewCSIMaxVolumeLimitPredicate(getFakeCSIPVInfo("csi-ebs", "csi-ebs"), getFakeCSIPVCInfo("csi-ebs"))
fits, reasons, err := pred(test.newPod, PredicateMetadata(test.newPod, nil), node)
if err != nil {
t.Errorf("Using allocatable [%s]%s: unexpected error: %v", test.filterName, test.test, err)
}
if !fits && !reflect.DeepEqual(reasons, expectedFailureReasons) {
t.Errorf("Using allocatable [%s]%s: unexpected failure reasons: %v, want: %v", test.filterName, test.test, reasons, expectedFailureReasons)
}
if fits != test.fits {
t.Errorf("Using allocatable [%s]%s: expected %v, got %v", test.filterName, test.test, test.fits, fits)
}
}
}
func getFakeCSIPVInfo(volumeName, driverName string) FakePersistentVolumeInfo {
return FakePersistentVolumeInfo{
{
ObjectMeta: metav1.ObjectMeta{Name: volumeName},
Spec: v1.PersistentVolumeSpec{
PersistentVolumeSource: v1.PersistentVolumeSource{
CSI: &v1.CSIPersistentVolumeSource{
Driver: driverName,
VolumeHandle: volumeName,
},
},
},
},
{
ObjectMeta: metav1.ObjectMeta{Name: volumeName + "-2"},
Spec: v1.PersistentVolumeSpec{
PersistentVolumeSource: v1.PersistentVolumeSource{
CSI: &v1.CSIPersistentVolumeSource{
Driver: driverName,
VolumeHandle: volumeName + "-2",
},
},
},
},
{
ObjectMeta: metav1.ObjectMeta{Name: volumeName + "-3"},
Spec: v1.PersistentVolumeSpec{
PersistentVolumeSource: v1.PersistentVolumeSource{
CSI: &v1.CSIPersistentVolumeSource{
Driver: driverName,
VolumeHandle: volumeName + "-3",
},
},
},
},
}
}
func getFakeCSIPVCInfo(volumeName string) FakePersistentVolumeClaimInfo {
return FakePersistentVolumeClaimInfo{
{
ObjectMeta: metav1.ObjectMeta{Name: volumeName},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: volumeName},
},
{
ObjectMeta: metav1.ObjectMeta{Name: volumeName + "-2"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: volumeName + "-2"},
},
{
ObjectMeta: metav1.ObjectMeta{Name: volumeName + "-3"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: volumeName + "-3"},
},
}
}

View File

@ -1,5 +1,5 @@
/*
Copyright 2016 The Kubernetes Authors.
Copyright 2018 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.
@ -14,23 +14,16 @@ See the License for the specific language governing permissions and
limitations under the License.
*/
package util
package predicates
import (
"k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"testing"
utilfeature "k8s.io/apiserver/pkg/util/feature"
utilfeaturetesting "k8s.io/apiserver/pkg/util/feature/testing"
_ "k8s.io/kubernetes/pkg/features"
)
// GetControllerRef gets pod's owner controller reference from a pod object.
func GetControllerRef(pod *v1.Pod) *metav1.OwnerReference {
if len(pod.OwnerReferences) == 0 {
return nil
}
for i := range pod.OwnerReferences {
ref := &pod.OwnerReferences[i]
if ref.Controller != nil && *ref.Controller {
return ref
}
}
return nil
func TestMain(m *testing.M) {
utilfeaturetesting.VerifyFeatureGatesUnchanged(utilfeature.DefaultFeatureGate, m.Run)
}

View File

@ -29,6 +29,7 @@ import (
utilfeature "k8s.io/apiserver/pkg/util/feature"
utilfeaturetesting "k8s.io/apiserver/pkg/util/feature/testing"
"k8s.io/kubernetes/pkg/features"
kubeletapis "k8s.io/kubernetes/pkg/kubelet/apis"
"k8s.io/kubernetes/pkg/scheduler/algorithm"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
volumeutil "k8s.io/kubernetes/pkg/volume/util"
@ -741,60 +742,12 @@ func TestVolumeCountConflicts(t *testing.T) {
},
}
pvInfo := func(filterName string) FakePersistentVolumeInfo {
return FakePersistentVolumeInfo{
{
ObjectMeta: metav1.ObjectMeta{Name: "some" + filterName + "Vol"},
Spec: v1.PersistentVolumeSpec{
PersistentVolumeSource: v1.PersistentVolumeSource{
AWSElasticBlockStore: &v1.AWSElasticBlockStoreVolumeSource{VolumeID: strings.ToLower(filterName) + "Vol"},
},
},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "someNon" + filterName + "Vol"},
Spec: v1.PersistentVolumeSpec{
PersistentVolumeSource: v1.PersistentVolumeSource{},
},
},
}
}
pvcInfo := func(filterName string) FakePersistentVolumeClaimInfo {
return FakePersistentVolumeClaimInfo{
{
ObjectMeta: metav1.ObjectMeta{Name: "some" + filterName + "Vol"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: "some" + filterName + "Vol"},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "someNon" + filterName + "Vol"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: "someNon" + filterName + "Vol"},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "pvcWithDeletedPV"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: "pvcWithDeletedPV"},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "anotherPVCWithDeletedPV"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: "anotherPVCWithDeletedPV"},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "unboundPVC"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: ""},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "anotherUnboundPVC"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: ""},
},
}
}
expectedFailureReasons := []algorithm.PredicateFailureReason{ErrMaxVolumeCountExceeded}
// running attachable predicate tests without feature gate and no limit present on nodes
for _, test := range tests {
os.Setenv(KubeMaxPDVols, strconv.Itoa(test.maxVols))
pred := NewMaxPDVolumeCountPredicate(test.filterName, pvInfo(test.filterName), pvcInfo(test.filterName))
pred := NewMaxPDVolumeCountPredicate(test.filterName, getFakePVInfo(test.filterName), getFakePVCInfo(test.filterName))
fits, reasons, err := pred(test.newPod, PredicateMetadata(test.newPod, nil), schedulercache.NewNodeInfo(test.existingPods...))
if err != nil {
t.Errorf("[%s]%s: unexpected error: %v", test.filterName, test.test, err)
@ -812,7 +765,7 @@ func TestVolumeCountConflicts(t *testing.T) {
// running attachable predicate tests with feature gate and limit present on nodes
for _, test := range tests {
node := getNodeWithPodAndVolumeLimits(test.existingPods, int64(test.maxVols), test.filterName)
pred := NewMaxPDVolumeCountPredicate(test.filterName, pvInfo(test.filterName), pvcInfo(test.filterName))
pred := NewMaxPDVolumeCountPredicate(test.filterName, getFakePVInfo(test.filterName), getFakePVCInfo(test.filterName))
fits, reasons, err := pred(test.newPod, PredicateMetadata(test.newPod, nil), node)
if err != nil {
t.Errorf("Using allocatable [%s]%s: unexpected error: %v", test.filterName, test.test, err)
@ -826,6 +779,122 @@ func TestVolumeCountConflicts(t *testing.T) {
}
}
func getFakePVInfo(filterName string) FakePersistentVolumeInfo {
return FakePersistentVolumeInfo{
{
ObjectMeta: metav1.ObjectMeta{Name: "some" + filterName + "Vol"},
Spec: v1.PersistentVolumeSpec{
PersistentVolumeSource: v1.PersistentVolumeSource{
AWSElasticBlockStore: &v1.AWSElasticBlockStoreVolumeSource{VolumeID: strings.ToLower(filterName) + "Vol"},
},
},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "someNon" + filterName + "Vol"},
Spec: v1.PersistentVolumeSpec{
PersistentVolumeSource: v1.PersistentVolumeSource{},
},
},
}
}
func getFakePVCInfo(filterName string) FakePersistentVolumeClaimInfo {
return FakePersistentVolumeClaimInfo{
{
ObjectMeta: metav1.ObjectMeta{Name: "some" + filterName + "Vol"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: "some" + filterName + "Vol"},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "someNon" + filterName + "Vol"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: "someNon" + filterName + "Vol"},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "pvcWithDeletedPV"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: "pvcWithDeletedPV"},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "anotherPVCWithDeletedPV"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: "anotherPVCWithDeletedPV"},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "unboundPVC"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: ""},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "anotherUnboundPVC"},
Spec: v1.PersistentVolumeClaimSpec{VolumeName: ""},
},
}
}
func TestMaxVolumeFuncM5(t *testing.T) {
node := &v1.Node{
ObjectMeta: metav1.ObjectMeta{
Name: "node-for-m5-instance",
Labels: map[string]string{
kubeletapis.LabelInstanceType: "m5.large",
},
},
}
os.Unsetenv(KubeMaxPDVols)
maxVolumeFunc := getMaxVolumeFunc(EBSVolumeFilterType)
maxVolume := maxVolumeFunc(node)
if maxVolume != volumeutil.DefaultMaxEBSNitroVolumeLimit {
t.Errorf("Expected max volume to be %d got %d", volumeutil.DefaultMaxEBSNitroVolumeLimit, maxVolume)
}
}
func TestMaxVolumeFuncT3(t *testing.T) {
node := &v1.Node{
ObjectMeta: metav1.ObjectMeta{
Name: "node-for-t3-instance",
Labels: map[string]string{
kubeletapis.LabelInstanceType: "t3.medium",
},
},
}
os.Unsetenv(KubeMaxPDVols)
maxVolumeFunc := getMaxVolumeFunc(EBSVolumeFilterType)
maxVolume := maxVolumeFunc(node)
if maxVolume != volumeutil.DefaultMaxEBSNitroVolumeLimit {
t.Errorf("Expected max volume to be %d got %d", volumeutil.DefaultMaxEBSNitroVolumeLimit, maxVolume)
}
}
func TestMaxVolumeFuncR5(t *testing.T) {
node := &v1.Node{
ObjectMeta: metav1.ObjectMeta{
Name: "node-for-r5-instance",
Labels: map[string]string{
kubeletapis.LabelInstanceType: "r5d.xlarge",
},
},
}
os.Unsetenv(KubeMaxPDVols)
maxVolumeFunc := getMaxVolumeFunc(EBSVolumeFilterType)
maxVolume := maxVolumeFunc(node)
if maxVolume != volumeutil.DefaultMaxEBSNitroVolumeLimit {
t.Errorf("Expected max volume to be %d got %d", volumeutil.DefaultMaxEBSNitroVolumeLimit, maxVolume)
}
}
func TestMaxVolumeFuncM4(t *testing.T) {
node := &v1.Node{
ObjectMeta: metav1.ObjectMeta{
Name: "node-for-m4-instance",
Labels: map[string]string{
kubeletapis.LabelInstanceType: "m4.2xlarge",
},
},
}
os.Unsetenv(KubeMaxPDVols)
maxVolumeFunc := getMaxVolumeFunc(EBSVolumeFilterType)
maxVolume := maxVolumeFunc(node)
if maxVolume != volumeutil.DefaultMaxEBSVolumes {
t.Errorf("Expected max volume to be %d got %d", volumeutil.DefaultMaxEBSVolumes, maxVolume)
}
}
func getNodeWithPodAndVolumeLimits(pods []*v1.Pod, limit int64, filter string) *schedulercache.NodeInfo {
nodeInfo := schedulercache.NewNodeInfo(pods...)
node := &v1.Node{
@ -849,6 +918,6 @@ func getVolumeLimitKey(filterType string) v1.ResourceName {
case AzureDiskVolumeFilterType:
return v1.ResourceName(volumeutil.AzureVolumeLimitKey)
default:
return ""
return v1.ResourceName(volumeutil.GetCSIAttachLimitKey(filterType))
}
}

View File

@ -17,10 +17,11 @@ limitations under the License.
package predicates
import (
"context"
"fmt"
"sync"
"github.com/golang/glog"
"k8s.io/klog"
"k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
@ -38,6 +39,12 @@ type PredicateMetadataFactory struct {
podLister algorithm.PodLister
}
// AntiAffinityTerm's topology key value used in predicate metadata
type topologyPair struct {
key string
value string
}
// Note that predicateMetadata and matchingPodAntiAffinityTerm need to be declared in the same file
// due to the way declarations are processed in predicate declaration unit tests.
type matchingPodAntiAffinityTerm struct {
@ -45,6 +52,17 @@ type matchingPodAntiAffinityTerm struct {
node *v1.Node
}
type podSet map[*v1.Pod]struct{}
type topologyPairSet map[topologyPair]struct{}
// topologyPairsMaps keeps topologyPairToAntiAffinityPods and antiAffinityPodToTopologyPairs in sync
// as they are the inverse of each others.
type topologyPairsMaps struct {
topologyPairToPods map[topologyPair]podSet
podToTopologyPairs map[string]topologyPairSet
}
// NOTE: When new fields are added/removed or logic is changed, please make sure that
// RemovePod, AddPod, and ShallowCopy functions are updated to work with the new changes.
type predicateMetadata struct {
@ -52,17 +70,17 @@ type predicateMetadata struct {
podBestEffort bool
podRequest *schedulercache.Resource
podPorts []*v1.ContainerPort
//key is a pod full name with the anti-affinity rules.
matchingAntiAffinityTerms map[string][]matchingPodAntiAffinityTerm
// A map of node name to a list of Pods on the node that can potentially match
// the affinity rules of the "pod".
nodeNameToMatchingAffinityPods map[string][]*v1.Pod
// A map of node name to a list of Pods on the node that can potentially match
// the anti-affinity rules of the "pod".
nodeNameToMatchingAntiAffinityPods map[string][]*v1.Pod
serviceAffinityInUse bool
serviceAffinityMatchingPodList []*v1.Pod
serviceAffinityMatchingPodServices []*v1.Service
topologyPairsAntiAffinityPodsMap *topologyPairsMaps
// A map of topology pairs to a list of Pods that can potentially match
// the affinity terms of the "pod" and its inverse.
topologyPairsPotentialAffinityPods *topologyPairsMaps
// A map of topology pairs to a list of Pods that can potentially match
// the anti-affinity terms of the "pod" and its inverse.
topologyPairsPotentialAntiAffinityPods *topologyPairsMaps
serviceAffinityInUse bool
serviceAffinityMatchingPodList []*v1.Pod
serviceAffinityMatchingPodServices []*v1.Service
// ignoredExtendedResources is a set of extended resource names that will
// be ignored in the PodFitsResources predicate.
//
@ -113,31 +131,74 @@ func (pfactory *PredicateMetadataFactory) GetMetadata(pod *v1.Pod, nodeNameToInf
if pod == nil {
return nil
}
matchingTerms, err := getMatchingAntiAffinityTerms(pod, nodeNameToInfoMap)
// existingPodAntiAffinityMap will be used later for efficient check on existing pods' anti-affinity
existingPodAntiAffinityMap, err := getTPMapMatchingExistingAntiAffinity(pod, nodeNameToInfoMap)
if err != nil {
return nil
}
affinityPods, antiAffinityPods, err := getPodsMatchingAffinity(pod, nodeNameToInfoMap)
// incomingPodAffinityMap will be used later for efficient check on incoming pod's affinity
// incomingPodAntiAffinityMap will be used later for efficient check on incoming pod's anti-affinity
incomingPodAffinityMap, incomingPodAntiAffinityMap, err := getTPMapMatchingIncomingAffinityAntiAffinity(pod, nodeNameToInfoMap)
if err != nil {
glog.Errorf("[predicate meta data generation] error finding pods that match affinity terms: %v", err)
klog.Errorf("[predicate meta data generation] error finding pods that match affinity terms: %v", err)
return nil
}
predicateMetadata := &predicateMetadata{
pod: pod,
podBestEffort: isPodBestEffort(pod),
podRequest: GetResourceRequest(pod),
podPorts: schedutil.GetContainerPorts(pod),
matchingAntiAffinityTerms: matchingTerms,
nodeNameToMatchingAffinityPods: affinityPods,
nodeNameToMatchingAntiAffinityPods: antiAffinityPods,
pod: pod,
podBestEffort: isPodBestEffort(pod),
podRequest: GetResourceRequest(pod),
podPorts: schedutil.GetContainerPorts(pod),
topologyPairsPotentialAffinityPods: incomingPodAffinityMap,
topologyPairsPotentialAntiAffinityPods: incomingPodAntiAffinityMap,
topologyPairsAntiAffinityPodsMap: existingPodAntiAffinityMap,
}
for predicateName, precomputeFunc := range predicateMetadataProducers {
glog.V(10).Infof("Precompute: %v", predicateName)
klog.V(10).Infof("Precompute: %v", predicateName)
precomputeFunc(predicateMetadata)
}
return predicateMetadata
}
// returns a pointer to a new topologyPairsMaps
func newTopologyPairsMaps() *topologyPairsMaps {
return &topologyPairsMaps{topologyPairToPods: make(map[topologyPair]podSet),
podToTopologyPairs: make(map[string]topologyPairSet)}
}
func (topologyPairsMaps *topologyPairsMaps) addTopologyPair(pair topologyPair, pod *v1.Pod) {
podFullName := schedutil.GetPodFullName(pod)
if topologyPairsMaps.topologyPairToPods[pair] == nil {
topologyPairsMaps.topologyPairToPods[pair] = make(map[*v1.Pod]struct{})
}
topologyPairsMaps.topologyPairToPods[pair][pod] = struct{}{}
if topologyPairsMaps.podToTopologyPairs[podFullName] == nil {
topologyPairsMaps.podToTopologyPairs[podFullName] = make(map[topologyPair]struct{})
}
topologyPairsMaps.podToTopologyPairs[podFullName][pair] = struct{}{}
}
func (topologyPairsMaps *topologyPairsMaps) removePod(deletedPod *v1.Pod) {
deletedPodFullName := schedutil.GetPodFullName(deletedPod)
for pair := range topologyPairsMaps.podToTopologyPairs[deletedPodFullName] {
delete(topologyPairsMaps.topologyPairToPods[pair], deletedPod)
if len(topologyPairsMaps.topologyPairToPods[pair]) == 0 {
delete(topologyPairsMaps.topologyPairToPods, pair)
}
}
delete(topologyPairsMaps.podToTopologyPairs, deletedPodFullName)
}
func (topologyPairsMaps *topologyPairsMaps) appendMaps(toAppend *topologyPairsMaps) {
if toAppend == nil {
return
}
for pair := range toAppend.topologyPairToPods {
for pod := range toAppend.topologyPairToPods[pair] {
topologyPairsMaps.addTopologyPair(pair, pod)
}
}
}
// RemovePod changes predicateMetadata assuming that the given `deletedPod` is
// deleted from the system.
func (meta *predicateMetadata) RemovePod(deletedPod *v1.Pod) error {
@ -145,35 +206,10 @@ func (meta *predicateMetadata) RemovePod(deletedPod *v1.Pod) error {
if deletedPodFullName == schedutil.GetPodFullName(meta.pod) {
return fmt.Errorf("deletedPod and meta.pod must not be the same")
}
// Delete any anti-affinity rule from the deletedPod.
delete(meta.matchingAntiAffinityTerms, deletedPodFullName)
// Delete pod from the matching affinity or anti-affinity pods if exists.
affinity := meta.pod.Spec.Affinity
podNodeName := deletedPod.Spec.NodeName
if affinity != nil && len(podNodeName) > 0 {
if affinity.PodAffinity != nil {
for i, p := range meta.nodeNameToMatchingAffinityPods[podNodeName] {
if p == deletedPod {
s := meta.nodeNameToMatchingAffinityPods[podNodeName]
s[i] = s[len(s)-1]
s = s[:len(s)-1]
meta.nodeNameToMatchingAffinityPods[podNodeName] = s
break
}
}
}
if affinity.PodAntiAffinity != nil {
for i, p := range meta.nodeNameToMatchingAntiAffinityPods[podNodeName] {
if p == deletedPod {
s := meta.nodeNameToMatchingAntiAffinityPods[podNodeName]
s[i] = s[len(s)-1]
s = s[:len(s)-1]
meta.nodeNameToMatchingAntiAffinityPods[podNodeName] = s
break
}
}
}
}
meta.topologyPairsAntiAffinityPodsMap.removePod(deletedPod)
// Delete pod from the matching affinity or anti-affinity topology pairs maps.
meta.topologyPairsPotentialAffinityPods.removePod(deletedPod)
meta.topologyPairsPotentialAntiAffinityPods.removePod(deletedPod)
// All pods in the serviceAffinityMatchingPodList are in the same namespace.
// So, if the namespace of the first one is not the same as the namespace of the
// deletedPod, we don't need to check the list, as deletedPod isn't in the list.
@ -203,46 +239,36 @@ func (meta *predicateMetadata) AddPod(addedPod *v1.Pod, nodeInfo *schedulercache
return fmt.Errorf("invalid node in nodeInfo")
}
// Add matching anti-affinity terms of the addedPod to the map.
podMatchingTerms, err := getMatchingAntiAffinityTermsOfExistingPod(meta.pod, addedPod, nodeInfo.Node())
topologyPairsMaps, err := getMatchingAntiAffinityTopologyPairsOfPod(meta.pod, addedPod, nodeInfo.Node())
if err != nil {
return err
}
if len(podMatchingTerms) > 0 {
existingTerms, found := meta.matchingAntiAffinityTerms[addedPodFullName]
if found {
meta.matchingAntiAffinityTerms[addedPodFullName] = append(existingTerms,
podMatchingTerms...)
} else {
meta.matchingAntiAffinityTerms[addedPodFullName] = podMatchingTerms
}
}
meta.topologyPairsAntiAffinityPodsMap.appendMaps(topologyPairsMaps)
// Add the pod to nodeNameToMatchingAffinityPods and nodeNameToMatchingAntiAffinityPods if needed.
affinity := meta.pod.Spec.Affinity
podNodeName := addedPod.Spec.NodeName
if affinity != nil && len(podNodeName) > 0 {
podNode := nodeInfo.Node()
// It is assumed that when the added pod matches affinity of the meta.pod, all the terms must match,
// this should be changed when the implementation of targetPodMatchesAffinityOfPod/podMatchesAffinityTermProperties
// is changed
if targetPodMatchesAffinityOfPod(meta.pod, addedPod) {
found := false
for _, p := range meta.nodeNameToMatchingAffinityPods[podNodeName] {
if p == addedPod {
found = true
break
affinityTerms := GetPodAffinityTerms(affinity.PodAffinity)
for _, term := range affinityTerms {
if topologyValue, ok := podNode.Labels[term.TopologyKey]; ok {
pair := topologyPair{key: term.TopologyKey, value: topologyValue}
meta.topologyPairsPotentialAffinityPods.addTopologyPair(pair, addedPod)
}
}
if !found {
meta.nodeNameToMatchingAffinityPods[podNodeName] = append(meta.nodeNameToMatchingAffinityPods[podNodeName], addedPod)
}
}
if targetPodMatchesAntiAffinityOfPod(meta.pod, addedPod) {
found := false
for _, p := range meta.nodeNameToMatchingAntiAffinityPods[podNodeName] {
if p == addedPod {
found = true
break
antiAffinityTerms := GetPodAntiAffinityTerms(affinity.PodAntiAffinity)
for _, term := range antiAffinityTerms {
if topologyValue, ok := podNode.Labels[term.TopologyKey]; ok {
pair := topologyPair{key: term.TopologyKey, value: topologyValue}
meta.topologyPairsPotentialAntiAffinityPods.addTopologyPair(pair, addedPod)
}
}
if !found {
meta.nodeNameToMatchingAntiAffinityPods[podNodeName] = append(meta.nodeNameToMatchingAntiAffinityPods[podNodeName], addedPod)
}
}
}
// If addedPod is in the same namespace as the meta.pod, update the list
@ -268,18 +294,12 @@ func (meta *predicateMetadata) ShallowCopy() algorithm.PredicateMetadata {
ignoredExtendedResources: meta.ignoredExtendedResources,
}
newPredMeta.podPorts = append([]*v1.ContainerPort(nil), meta.podPorts...)
newPredMeta.matchingAntiAffinityTerms = map[string][]matchingPodAntiAffinityTerm{}
for k, v := range meta.matchingAntiAffinityTerms {
newPredMeta.matchingAntiAffinityTerms[k] = append([]matchingPodAntiAffinityTerm(nil), v...)
}
newPredMeta.nodeNameToMatchingAffinityPods = make(map[string][]*v1.Pod)
for k, v := range meta.nodeNameToMatchingAffinityPods {
newPredMeta.nodeNameToMatchingAffinityPods[k] = append([]*v1.Pod(nil), v...)
}
newPredMeta.nodeNameToMatchingAntiAffinityPods = make(map[string][]*v1.Pod)
for k, v := range meta.nodeNameToMatchingAntiAffinityPods {
newPredMeta.nodeNameToMatchingAntiAffinityPods[k] = append([]*v1.Pod(nil), v...)
}
newPredMeta.topologyPairsPotentialAffinityPods = newTopologyPairsMaps()
newPredMeta.topologyPairsPotentialAffinityPods.appendMaps(meta.topologyPairsPotentialAffinityPods)
newPredMeta.topologyPairsPotentialAntiAffinityPods = newTopologyPairsMaps()
newPredMeta.topologyPairsPotentialAntiAffinityPods.appendMaps(meta.topologyPairsPotentialAntiAffinityPods)
newPredMeta.topologyPairsAntiAffinityPodsMap = newTopologyPairsMaps()
newPredMeta.topologyPairsAntiAffinityPodsMap.appendMaps(meta.topologyPairsAntiAffinityPodsMap)
newPredMeta.serviceAffinityMatchingPodServices = append([]*v1.Service(nil),
meta.serviceAffinityMatchingPodServices...)
newPredMeta.serviceAffinityMatchingPodList = append([]*v1.Pod(nil),
@ -310,8 +330,8 @@ func getAffinityTermProperties(pod *v1.Pod, terms []v1.PodAffinityTerm) (propert
return properties, nil
}
// podMatchesAffinityTermProperties return true IFF the given pod matches all the given properties.
func podMatchesAffinityTermProperties(pod *v1.Pod, properties []*affinityTermProperties) bool {
// podMatchesAllAffinityTermProperties returns true IFF the given pod matches all the given properties.
func podMatchesAllAffinityTermProperties(pod *v1.Pod, properties []*affinityTermProperties) bool {
if len(properties) == 0 {
return false
}
@ -323,34 +343,93 @@ func podMatchesAffinityTermProperties(pod *v1.Pod, properties []*affinityTermPro
return true
}
// getPodsMatchingAffinity finds existing Pods that match affinity terms of the given "pod".
// It ignores topology. It returns a set of Pods that are checked later by the affinity
// predicate. With this set of pods available, the affinity predicate does not
// need to check all the pods in the cluster.
func getPodsMatchingAffinity(pod *v1.Pod, nodeInfoMap map[string]*schedulercache.NodeInfo) (affinityPods map[string][]*v1.Pod, antiAffinityPods map[string][]*v1.Pod, err error) {
allNodeNames := make([]string, 0, len(nodeInfoMap))
affinity := pod.Spec.Affinity
if affinity == nil || (affinity.PodAffinity == nil && affinity.PodAntiAffinity == nil) {
return nil, nil, nil
// podMatchesAnyAffinityTermProperties returns true if the given pod matches any given property.
func podMatchesAnyAffinityTermProperties(pod *v1.Pod, properties []*affinityTermProperties) bool {
if len(properties) == 0 {
return false
}
for _, property := range properties {
if priorityutil.PodMatchesTermsNamespaceAndSelector(pod, property.namespaces, property.selector) {
return true
}
}
return false
}
// getTPMapMatchingExistingAntiAffinity calculates the following for each existing pod on each node:
// (1) Whether it has PodAntiAffinity
// (2) Whether any AffinityTerm matches the incoming pod
func getTPMapMatchingExistingAntiAffinity(pod *v1.Pod, nodeInfoMap map[string]*schedulercache.NodeInfo) (*topologyPairsMaps, error) {
allNodeNames := make([]string, 0, len(nodeInfoMap))
for name := range nodeInfoMap {
allNodeNames = append(allNodeNames, name)
}
var lock sync.Mutex
var firstError error
affinityPods = make(map[string][]*v1.Pod)
antiAffinityPods = make(map[string][]*v1.Pod)
appendResult := func(nodeName string, affPods, antiAffPods []*v1.Pod) {
topologyMaps := newTopologyPairsMaps()
appendTopologyPairsMaps := func(toAppend *topologyPairsMaps) {
lock.Lock()
defer lock.Unlock()
if len(affPods) > 0 {
affinityPods[nodeName] = affPods
topologyMaps.appendMaps(toAppend)
}
catchError := func(err error) {
lock.Lock()
defer lock.Unlock()
if firstError == nil {
firstError = err
}
if len(antiAffPods) > 0 {
antiAffinityPods[nodeName] = antiAffPods
}
processNode := func(i int) {
nodeInfo := nodeInfoMap[allNodeNames[i]]
node := nodeInfo.Node()
if node == nil {
catchError(fmt.Errorf("node not found"))
return
}
for _, existingPod := range nodeInfo.PodsWithAffinity() {
existingPodTopologyMaps, err := getMatchingAntiAffinityTopologyPairsOfPod(pod, existingPod, node)
if err != nil {
catchError(err)
return
}
appendTopologyPairsMaps(existingPodTopologyMaps)
}
}
workqueue.ParallelizeUntil(context.TODO(), 16, len(allNodeNames), processNode)
return topologyMaps, firstError
}
// getTPMapMatchingIncomingAffinityAntiAffinity finds existing Pods that match affinity terms of the given "pod".
// It returns a topologyPairsMaps that are checked later by the affinity
// predicate. With this topologyPairsMaps available, the affinity predicate does not
// need to check all the pods in the cluster.
func getTPMapMatchingIncomingAffinityAntiAffinity(pod *v1.Pod, nodeInfoMap map[string]*schedulercache.NodeInfo) (topologyPairsAffinityPodsMaps *topologyPairsMaps, topologyPairsAntiAffinityPodsMaps *topologyPairsMaps, err error) {
affinity := pod.Spec.Affinity
if affinity == nil || (affinity.PodAffinity == nil && affinity.PodAntiAffinity == nil) {
return newTopologyPairsMaps(), newTopologyPairsMaps(), nil
}
allNodeNames := make([]string, 0, len(nodeInfoMap))
for name := range nodeInfoMap {
allNodeNames = append(allNodeNames, name)
}
var lock sync.Mutex
var firstError error
topologyPairsAffinityPodsMaps = newTopologyPairsMaps()
topologyPairsAntiAffinityPodsMaps = newTopologyPairsMaps()
appendResult := func(nodeName string, nodeTopologyPairsAffinityPodsMaps, nodeTopologyPairsAntiAffinityPodsMaps *topologyPairsMaps) {
lock.Lock()
defer lock.Unlock()
if len(nodeTopologyPairsAffinityPodsMaps.topologyPairToPods) > 0 {
topologyPairsAffinityPodsMaps.appendMaps(nodeTopologyPairsAffinityPodsMaps)
}
if len(nodeTopologyPairsAntiAffinityPodsMaps.topologyPairToPods) > 0 {
topologyPairsAntiAffinityPodsMaps.appendMaps(nodeTopologyPairsAntiAffinityPodsMaps)
}
}
@ -362,14 +441,12 @@ func getPodsMatchingAffinity(pod *v1.Pod, nodeInfoMap map[string]*schedulercache
}
}
affinityProperties, err := getAffinityTermProperties(pod, GetPodAffinityTerms(affinity.PodAffinity))
if err != nil {
return nil, nil, err
}
antiAffinityProperties, err := getAffinityTermProperties(pod, GetPodAntiAffinityTerms(affinity.PodAntiAffinity))
affinityTerms := GetPodAffinityTerms(affinity.PodAffinity)
affinityProperties, err := getAffinityTermProperties(pod, affinityTerms)
if err != nil {
return nil, nil, err
}
antiAffinityTerms := GetPodAntiAffinityTerms(affinity.PodAntiAffinity)
processNode := func(i int) {
nodeInfo := nodeInfoMap[allNodeNames[i]]
@ -378,28 +455,44 @@ func getPodsMatchingAffinity(pod *v1.Pod, nodeInfoMap map[string]*schedulercache
catchError(fmt.Errorf("nodeInfo.Node is nil"))
return
}
affPods := make([]*v1.Pod, 0, len(nodeInfo.Pods()))
antiAffPods := make([]*v1.Pod, 0, len(nodeInfo.Pods()))
nodeTopologyPairsAffinityPodsMaps := newTopologyPairsMaps()
nodeTopologyPairsAntiAffinityPodsMaps := newTopologyPairsMaps()
for _, existingPod := range nodeInfo.Pods() {
// Check affinity properties.
if podMatchesAffinityTermProperties(existingPod, affinityProperties) {
affPods = append(affPods, existingPod)
if podMatchesAllAffinityTermProperties(existingPod, affinityProperties) {
for _, term := range affinityTerms {
if topologyValue, ok := node.Labels[term.TopologyKey]; ok {
pair := topologyPair{key: term.TopologyKey, value: topologyValue}
nodeTopologyPairsAffinityPodsMaps.addTopologyPair(pair, existingPod)
}
}
}
// Check anti-affinity properties.
if podMatchesAffinityTermProperties(existingPod, antiAffinityProperties) {
antiAffPods = append(antiAffPods, existingPod)
for _, term := range antiAffinityTerms {
namespaces := priorityutil.GetNamespacesFromPodAffinityTerm(pod, &term)
selector, err := metav1.LabelSelectorAsSelector(term.LabelSelector)
if err != nil {
catchError(err)
return
}
if priorityutil.PodMatchesTermsNamespaceAndSelector(existingPod, namespaces, selector) {
if topologyValue, ok := node.Labels[term.TopologyKey]; ok {
pair := topologyPair{key: term.TopologyKey, value: topologyValue}
nodeTopologyPairsAntiAffinityPodsMaps.addTopologyPair(pair, existingPod)
}
}
}
}
if len(antiAffPods) > 0 || len(affPods) > 0 {
appendResult(node.Name, affPods, antiAffPods)
if len(nodeTopologyPairsAffinityPodsMaps.topologyPairToPods) > 0 || len(nodeTopologyPairsAntiAffinityPodsMaps.topologyPairToPods) > 0 {
appendResult(node.Name, nodeTopologyPairsAffinityPodsMaps, nodeTopologyPairsAntiAffinityPodsMaps)
}
}
workqueue.Parallelize(16, len(allNodeNames), processNode)
return affinityPods, antiAffinityPods, firstError
workqueue.ParallelizeUntil(context.TODO(), 16, len(allNodeNames), processNode)
return topologyPairsAffinityPodsMaps, topologyPairsAntiAffinityPodsMaps, firstError
}
// podMatchesAffinity returns true if "targetPod" matches any affinity rule of
// "pod". Similar to getPodsMatchingAffinity, this function does not check topology.
// targetPodMatchesAffinityOfPod returns true if "targetPod" matches ALL affinity terms of
// "pod". This function does not check topology.
// So, whether the targetPod actually matches or not needs further checks for a specific
// node.
func targetPodMatchesAffinityOfPod(pod, targetPod *v1.Pod) bool {
@ -409,14 +502,14 @@ func targetPodMatchesAffinityOfPod(pod, targetPod *v1.Pod) bool {
}
affinityProperties, err := getAffinityTermProperties(pod, GetPodAffinityTerms(affinity.PodAffinity))
if err != nil {
glog.Errorf("error in getting affinity properties of Pod %v", pod.Name)
klog.Errorf("error in getting affinity properties of Pod %v", pod.Name)
return false
}
return podMatchesAffinityTermProperties(targetPod, affinityProperties)
return podMatchesAllAffinityTermProperties(targetPod, affinityProperties)
}
// targetPodMatchesAntiAffinityOfPod returns true if "targetPod" matches any anti-affinity
// rule of "pod". Similar to getPodsMatchingAffinity, this function does not check topology.
// targetPodMatchesAntiAffinityOfPod returns true if "targetPod" matches ANY anti-affinity
// term of "pod". This function does not check topology.
// So, whether the targetPod actually matches or not needs further checks for a specific
// node.
func targetPodMatchesAntiAffinityOfPod(pod, targetPod *v1.Pod) bool {
@ -426,8 +519,8 @@ func targetPodMatchesAntiAffinityOfPod(pod, targetPod *v1.Pod) bool {
}
properties, err := getAffinityTermProperties(pod, GetPodAntiAffinityTerms(affinity.PodAntiAffinity))
if err != nil {
glog.Errorf("error in getting anti-affinity properties of Pod %v", pod.Name)
klog.Errorf("error in getting anti-affinity properties of Pod %v", pod.Name)
return false
}
return podMatchesAffinityTermProperties(targetPod, properties)
return podMatchesAnyAffinityTermProperties(targetPod, properties)
}

View File

@ -28,42 +28,6 @@ import (
schedulertesting "k8s.io/kubernetes/pkg/scheduler/testing"
)
// sortableAntiAffinityTerms lets us to sort anti-affinity terms.
type sortableAntiAffinityTerms []matchingPodAntiAffinityTerm
// Less establishes some ordering between two matchingPodAntiAffinityTerms for
// sorting.
func (s sortableAntiAffinityTerms) Less(i, j int) bool {
t1, t2 := s[i], s[j]
if t1.node.Name != t2.node.Name {
return t1.node.Name < t2.node.Name
}
if len(t1.term.Namespaces) != len(t2.term.Namespaces) {
return len(t1.term.Namespaces) < len(t2.term.Namespaces)
}
if t1.term.TopologyKey != t2.term.TopologyKey {
return t1.term.TopologyKey < t2.term.TopologyKey
}
if len(t1.term.LabelSelector.MatchLabels) != len(t2.term.LabelSelector.MatchLabels) {
return len(t1.term.LabelSelector.MatchLabels) < len(t2.term.LabelSelector.MatchLabels)
}
return false
}
func (s sortableAntiAffinityTerms) Len() int { return len(s) }
func (s sortableAntiAffinityTerms) Swap(i, j int) {
s[i], s[j] = s[j], s[i]
}
var _ = sort.Interface(sortableAntiAffinityTerms{})
func sortAntiAffinityTerms(terms map[string][]matchingPodAntiAffinityTerm) {
for k, v := range terms {
sortableTerms := sortableAntiAffinityTerms(v)
sort.Sort(sortableTerms)
terms[k] = sortableTerms
}
}
// sortablePods lets us to sort pods.
type sortablePods []*v1.Pod
@ -88,13 +52,6 @@ func (s sortableServices) Swap(i, j int) { s[i], s[j] = s[j], s[i] }
var _ = sort.Interface(&sortableServices{})
func sortNodePodMap(np map[string][]*v1.Pod) {
for _, pl := range np {
sortablePods := sortablePods(pl)
sort.Sort(sortablePods)
}
}
// predicateMetadataEquivalent returns true if the two metadata are equivalent.
// Note: this function does not compare podRequest.
func predicateMetadataEquivalent(meta1, meta2 *predicateMetadata) error {
@ -113,20 +70,19 @@ func predicateMetadataEquivalent(meta1, meta2 *predicateMetadata) error {
for !reflect.DeepEqual(meta1.podPorts, meta2.podPorts) {
return fmt.Errorf("podPorts are not equal")
}
sortAntiAffinityTerms(meta1.matchingAntiAffinityTerms)
sortAntiAffinityTerms(meta2.matchingAntiAffinityTerms)
if !reflect.DeepEqual(meta1.matchingAntiAffinityTerms, meta2.matchingAntiAffinityTerms) {
return fmt.Errorf("matchingAntiAffinityTerms are not euqal")
if !reflect.DeepEqual(meta1.topologyPairsPotentialAffinityPods, meta2.topologyPairsPotentialAffinityPods) {
return fmt.Errorf("topologyPairsPotentialAffinityPods are not equal")
}
sortNodePodMap(meta1.nodeNameToMatchingAffinityPods)
sortNodePodMap(meta2.nodeNameToMatchingAffinityPods)
if !reflect.DeepEqual(meta1.nodeNameToMatchingAffinityPods, meta2.nodeNameToMatchingAffinityPods) {
return fmt.Errorf("nodeNameToMatchingAffinityPods are not euqal")
if !reflect.DeepEqual(meta1.topologyPairsPotentialAntiAffinityPods, meta2.topologyPairsPotentialAntiAffinityPods) {
return fmt.Errorf("topologyPairsPotentialAntiAffinityPods are not equal")
}
sortNodePodMap(meta1.nodeNameToMatchingAntiAffinityPods)
sortNodePodMap(meta2.nodeNameToMatchingAntiAffinityPods)
if !reflect.DeepEqual(meta1.nodeNameToMatchingAntiAffinityPods, meta2.nodeNameToMatchingAntiAffinityPods) {
return fmt.Errorf("nodeNameToMatchingAntiAffinityPods are not euqal")
if !reflect.DeepEqual(meta1.topologyPairsAntiAffinityPodsMap.podToTopologyPairs,
meta2.topologyPairsAntiAffinityPodsMap.podToTopologyPairs) {
return fmt.Errorf("topologyPairsAntiAffinityPodsMap.podToTopologyPairs are not equal")
}
if !reflect.DeepEqual(meta1.topologyPairsAntiAffinityPodsMap.topologyPairToPods,
meta2.topologyPairsAntiAffinityPodsMap.topologyPairToPods) {
return fmt.Errorf("topologyPairsAntiAffinityPodsMap.topologyPairToPods are not equal")
}
if meta1.serviceAffinityInUse {
sortablePods1 := sortablePods(meta1.serviceAffinityMatchingPodList)
@ -236,7 +192,7 @@ func TestPredicateMetadata_AddRemovePod(t *testing.T) {
}
tests := []struct {
description string
name string
pendingPod *v1.Pod
addedPod *v1.Pod
existingPods []*v1.Pod
@ -244,7 +200,7 @@ func TestPredicateMetadata_AddRemovePod(t *testing.T) {
services []*v1.Service
}{
{
description: "no anti-affinity or service affinity exist",
name: "no anti-affinity or service affinity exist",
pendingPod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "pending", Labels: selector1},
},
@ -267,7 +223,7 @@ func TestPredicateMetadata_AddRemovePod(t *testing.T) {
},
},
{
description: "metadata anti-affinity terms are updated correctly after adding and removing a pod",
name: "metadata anti-affinity terms are updated correctly after adding and removing a pod",
pendingPod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "pending", Labels: selector1},
},
@ -300,7 +256,7 @@ func TestPredicateMetadata_AddRemovePod(t *testing.T) {
},
},
{
description: "metadata service-affinity data are updated correctly after adding and removing a pod",
name: "metadata service-affinity data are updated correctly after adding and removing a pod",
pendingPod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "pending", Labels: selector1},
},
@ -324,7 +280,7 @@ func TestPredicateMetadata_AddRemovePod(t *testing.T) {
},
},
{
description: "metadata anti-affinity terms and service affinity data are updated correctly after adding and removing a pod",
name: "metadata anti-affinity terms and service affinity data are updated correctly after adding and removing a pod",
pendingPod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "pending", Labels: selector1},
},
@ -358,7 +314,7 @@ func TestPredicateMetadata_AddRemovePod(t *testing.T) {
},
},
{
description: "metadata matching pod affinity and anti-affinity are updated correctly after adding and removing a pod",
name: "metadata matching pod affinity and anti-affinity are updated correctly after adding and removing a pod",
pendingPod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "pending", Labels: selector1},
},
@ -395,44 +351,46 @@ func TestPredicateMetadata_AddRemovePod(t *testing.T) {
}
for _, test := range tests {
allPodLister := schedulertesting.FakePodLister(append(test.existingPods, test.addedPod))
// getMeta creates predicate meta data given the list of pods.
getMeta := func(lister schedulertesting.FakePodLister) (*predicateMetadata, map[string]*schedulercache.NodeInfo) {
nodeInfoMap := schedulercache.CreateNodeNameToInfoMap(lister, test.nodes)
// nodeList is a list of non-pointer nodes to feed to FakeNodeListInfo.
nodeList := []v1.Node{}
for _, n := range test.nodes {
nodeList = append(nodeList, *n)
t.Run(test.name, func(t *testing.T) {
allPodLister := schedulertesting.FakePodLister(append(test.existingPods, test.addedPod))
// getMeta creates predicate meta data given the list of pods.
getMeta := func(lister schedulertesting.FakePodLister) (*predicateMetadata, map[string]*schedulercache.NodeInfo) {
nodeInfoMap := schedulercache.CreateNodeNameToInfoMap(lister, test.nodes)
// nodeList is a list of non-pointer nodes to feed to FakeNodeListInfo.
nodeList := []v1.Node{}
for _, n := range test.nodes {
nodeList = append(nodeList, *n)
}
_, precompute := NewServiceAffinityPredicate(lister, schedulertesting.FakeServiceLister(test.services), FakeNodeListInfo(nodeList), nil)
RegisterPredicateMetadataProducer("ServiceAffinityMetaProducer", precompute)
pmf := PredicateMetadataFactory{lister}
meta := pmf.GetMetadata(test.pendingPod, nodeInfoMap)
return meta.(*predicateMetadata), nodeInfoMap
}
_, precompute := NewServiceAffinityPredicate(lister, schedulertesting.FakeServiceLister(test.services), FakeNodeListInfo(nodeList), nil)
RegisterPredicateMetadataProducer("ServiceAffinityMetaProducer", precompute)
pmf := PredicateMetadataFactory{lister}
meta := pmf.GetMetadata(test.pendingPod, nodeInfoMap)
return meta.(*predicateMetadata), nodeInfoMap
}
// allPodsMeta is meta data produced when all pods, including test.addedPod
// are given to the metadata producer.
allPodsMeta, _ := getMeta(allPodLister)
// existingPodsMeta1 is meta data produced for test.existingPods (without test.addedPod).
existingPodsMeta1, nodeInfoMap := getMeta(schedulertesting.FakePodLister(test.existingPods))
// Add test.addedPod to existingPodsMeta1 and make sure meta is equal to allPodsMeta
nodeInfo := nodeInfoMap[test.addedPod.Spec.NodeName]
if err := existingPodsMeta1.AddPod(test.addedPod, nodeInfo); err != nil {
t.Errorf("test [%v]: error adding pod to meta: %v", test.description, err)
}
if err := predicateMetadataEquivalent(allPodsMeta, existingPodsMeta1); err != nil {
t.Errorf("test [%v]: meta data are not equivalent: %v", test.description, err)
}
// Remove the added pod and from existingPodsMeta1 an make sure it is equal
// to meta generated for existing pods.
existingPodsMeta2, _ := getMeta(schedulertesting.FakePodLister(test.existingPods))
if err := existingPodsMeta1.RemovePod(test.addedPod); err != nil {
t.Errorf("test [%v]: error removing pod from meta: %v", test.description, err)
}
if err := predicateMetadataEquivalent(existingPodsMeta1, existingPodsMeta2); err != nil {
t.Errorf("test [%v]: meta data are not equivalent: %v", test.description, err)
}
// allPodsMeta is meta data produced when all pods, including test.addedPod
// are given to the metadata producer.
allPodsMeta, _ := getMeta(allPodLister)
// existingPodsMeta1 is meta data produced for test.existingPods (without test.addedPod).
existingPodsMeta1, nodeInfoMap := getMeta(schedulertesting.FakePodLister(test.existingPods))
// Add test.addedPod to existingPodsMeta1 and make sure meta is equal to allPodsMeta
nodeInfo := nodeInfoMap[test.addedPod.Spec.NodeName]
if err := existingPodsMeta1.AddPod(test.addedPod, nodeInfo); err != nil {
t.Errorf("error adding pod to meta: %v", err)
}
if err := predicateMetadataEquivalent(allPodsMeta, existingPodsMeta1); err != nil {
t.Errorf("meta data are not equivalent: %v", err)
}
// Remove the added pod and from existingPodsMeta1 an make sure it is equal
// to meta generated for existing pods.
existingPodsMeta2, _ := getMeta(schedulertesting.FakePodLister(test.existingPods))
if err := existingPodsMeta1.RemovePod(test.addedPod); err != nil {
t.Errorf("error removing pod from meta: %v", err)
}
if err := predicateMetadataEquivalent(existingPodsMeta1, existingPodsMeta2); err != nil {
t.Errorf("meta data are not equivalent: %v", err)
}
})
}
}
@ -463,52 +421,93 @@ func TestPredicateMetadata_ShallowCopy(t *testing.T) {
HostIP: "1.2.3.4",
},
},
matchingAntiAffinityTerms: map[string][]matchingPodAntiAffinityTerm{
"term1": {
{
term: &v1.PodAffinityTerm{TopologyKey: "node"},
node: &v1.Node{
ObjectMeta: metav1.ObjectMeta{Name: "machine1"},
},
topologyPairsAntiAffinityPodsMap: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "name", value: "machine1"}: {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p2", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeC"},
}: struct{}{},
},
{key: "name", value: "machine2"}: {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p1", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeA"},
}: struct{}{},
},
},
podToTopologyPairs: map[string]topologyPairSet{
"p2_": {
topologyPair{key: "name", value: "machine1"}: struct{}{},
},
"p1_": {
topologyPair{key: "name", value: "machine2"}: struct{}{},
},
},
},
nodeNameToMatchingAffinityPods: map[string][]*v1.Pod{
"nodeA": {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p1", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeA"},
topologyPairsPotentialAffinityPods: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "name", value: "nodeA"}: {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p1", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeA"},
}: struct{}{},
},
{key: "name", value: "nodeC"}: {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p2"},
Spec: v1.PodSpec{
NodeName: "nodeC",
},
}: struct{}{},
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p6", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeC"},
}: struct{}{},
},
},
"nodeC": {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p2"},
Spec: v1.PodSpec{
NodeName: "nodeC",
},
podToTopologyPairs: map[string]topologyPairSet{
"p1_": {
topologyPair{key: "name", value: "nodeA"}: struct{}{},
},
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p6", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeC"},
"p2_": {
topologyPair{key: "name", value: "nodeC"}: struct{}{},
},
"p6_": {
topologyPair{key: "name", value: "nodeC"}: struct{}{},
},
},
},
nodeNameToMatchingAntiAffinityPods: map[string][]*v1.Pod{
"nodeN": {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p1", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeN"},
topologyPairsPotentialAntiAffinityPods: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "name", value: "nodeN"}: {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p1", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeN"},
}: struct{}{},
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p2"},
Spec: v1.PodSpec{
NodeName: "nodeM",
},
}: struct{}{},
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p3"},
Spec: v1.PodSpec{
NodeName: "nodeM",
},
}: struct{}{},
},
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p2"},
Spec: v1.PodSpec{
NodeName: "nodeM",
},
},
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p3"},
Spec: v1.PodSpec{
NodeName: "nodeM",
},
{key: "name", value: "nodeM"}: {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p6", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeM"},
}: struct{}{},
},
},
"nodeM": {
&v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p6", Labels: selector1},
Spec: v1.PodSpec{NodeName: "nodeM"},
podToTopologyPairs: map[string]topologyPairSet{
"p1_": {
topologyPair{key: "name", value: "nodeN"}: struct{}{},
},
"p2_": {
topologyPair{key: "name", value: "nodeN"}: struct{}{},
},
"p3_": {
topologyPair{key: "name", value: "nodeN"}: struct{}{},
},
"p6_": {
topologyPair{key: "name", value: "nodeM"}: struct{}{},
},
},
},
@ -526,3 +525,269 @@ func TestPredicateMetadata_ShallowCopy(t *testing.T) {
t.Errorf("Copy is not equal to source!")
}
}
// TestGetTPMapMatchingIncomingAffinityAntiAffinity tests against method getTPMapMatchingIncomingAffinityAntiAffinity
// on Anti Affinity cases
func TestGetTPMapMatchingIncomingAffinityAntiAffinity(t *testing.T) {
newPodAffinityTerms := func(keys ...string) []v1.PodAffinityTerm {
var terms []v1.PodAffinityTerm
for _, key := range keys {
terms = append(terms, v1.PodAffinityTerm{
LabelSelector: &metav1.LabelSelector{
MatchExpressions: []metav1.LabelSelectorRequirement{
{
Key: key,
Operator: metav1.LabelSelectorOpExists,
},
},
},
TopologyKey: "hostname",
})
}
return terms
}
newPod := func(labels ...string) *v1.Pod {
labelMap := make(map[string]string)
for _, l := range labels {
labelMap[l] = ""
}
return &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "normal", Labels: labelMap},
Spec: v1.PodSpec{NodeName: "nodeA"},
}
}
normalPodA := newPod("aaa")
normalPodB := newPod("bbb")
normalPodAB := newPod("aaa", "bbb")
nodeA := &v1.Node{ObjectMeta: metav1.ObjectMeta{Name: "nodeA", Labels: map[string]string{"hostname": "nodeA"}}}
tests := []struct {
name string
existingPods []*v1.Pod
nodes []*v1.Node
pod *v1.Pod
wantAffinityPodsMaps *topologyPairsMaps
wantAntiAffinityPodsMaps *topologyPairsMaps
wantErr bool
}{
{
name: "nil test",
nodes: []*v1.Node{nodeA},
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "aaa-normal"},
},
wantAffinityPodsMaps: newTopologyPairsMaps(),
wantAntiAffinityPodsMaps: newTopologyPairsMaps(),
},
{
name: "incoming pod without affinity/anti-affinity causes a no-op",
existingPods: []*v1.Pod{normalPodA},
nodes: []*v1.Node{nodeA},
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "aaa-normal"},
},
wantAffinityPodsMaps: newTopologyPairsMaps(),
wantAntiAffinityPodsMaps: newTopologyPairsMaps(),
},
{
name: "no pod has label that violates incoming pod's affinity and anti-affinity",
existingPods: []*v1.Pod{normalPodB},
nodes: []*v1.Node{nodeA},
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "aaa-anti"},
Spec: v1.PodSpec{
Affinity: &v1.Affinity{
PodAffinity: &v1.PodAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa"),
},
PodAntiAffinity: &v1.PodAntiAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa"),
},
},
},
},
wantAffinityPodsMaps: newTopologyPairsMaps(),
wantAntiAffinityPodsMaps: newTopologyPairsMaps(),
},
{
name: "existing pod matches incoming pod's affinity and anti-affinity - single term case",
existingPods: []*v1.Pod{normalPodA},
nodes: []*v1.Node{nodeA},
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "affi-antiaffi"},
Spec: v1.PodSpec{
Affinity: &v1.Affinity{
PodAffinity: &v1.PodAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa"),
},
PodAntiAffinity: &v1.PodAntiAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa"),
},
},
},
},
wantAffinityPodsMaps: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "hostname", value: "nodeA"}: {normalPodA: struct{}{}},
},
podToTopologyPairs: map[string]topologyPairSet{
"normal_": {
topologyPair{key: "hostname", value: "nodeA"}: struct{}{},
},
},
},
wantAntiAffinityPodsMaps: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "hostname", value: "nodeA"}: {normalPodA: struct{}{}},
},
podToTopologyPairs: map[string]topologyPairSet{
"normal_": {
topologyPair{key: "hostname", value: "nodeA"}: struct{}{},
},
},
},
},
{
name: "existing pod matches incoming pod's affinity and anti-affinity - mutiple terms case",
existingPods: []*v1.Pod{normalPodAB},
nodes: []*v1.Node{nodeA},
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "affi-antiaffi"},
Spec: v1.PodSpec{
Affinity: &v1.Affinity{
PodAffinity: &v1.PodAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa", "bbb"),
},
PodAntiAffinity: &v1.PodAntiAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa"),
},
},
},
},
wantAffinityPodsMaps: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "hostname", value: "nodeA"}: {normalPodAB: struct{}{}},
},
podToTopologyPairs: map[string]topologyPairSet{
"normal_": {
topologyPair{key: "hostname", value: "nodeA"}: struct{}{},
},
},
},
wantAntiAffinityPodsMaps: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "hostname", value: "nodeA"}: {normalPodAB: struct{}{}},
},
podToTopologyPairs: map[string]topologyPairSet{
"normal_": {
topologyPair{key: "hostname", value: "nodeA"}: struct{}{},
},
},
},
},
{
name: "existing pod not match incoming pod's affinity but matches anti-affinity",
existingPods: []*v1.Pod{normalPodA},
nodes: []*v1.Node{nodeA},
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "affi-antiaffi"},
Spec: v1.PodSpec{
Affinity: &v1.Affinity{
PodAffinity: &v1.PodAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa", "bbb"),
},
PodAntiAffinity: &v1.PodAntiAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa", "bbb"),
},
},
},
},
wantAffinityPodsMaps: newTopologyPairsMaps(),
wantAntiAffinityPodsMaps: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "hostname", value: "nodeA"}: {normalPodA: struct{}{}},
},
podToTopologyPairs: map[string]topologyPairSet{
"normal_": {
topologyPair{key: "hostname", value: "nodeA"}: struct{}{},
},
},
},
},
{
name: "incoming pod's anti-affinity has more than one term - existing pod violates partial term - case 1",
existingPods: []*v1.Pod{normalPodAB},
nodes: []*v1.Node{nodeA},
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "anaffi-antiaffiti"},
Spec: v1.PodSpec{
Affinity: &v1.Affinity{
PodAffinity: &v1.PodAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa", "ccc"),
},
PodAntiAffinity: &v1.PodAntiAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa", "ccc"),
},
},
},
},
wantAffinityPodsMaps: newTopologyPairsMaps(),
wantAntiAffinityPodsMaps: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "hostname", value: "nodeA"}: {normalPodAB: struct{}{}},
},
podToTopologyPairs: map[string]topologyPairSet{
"normal_": {
topologyPair{key: "hostname", value: "nodeA"}: struct{}{},
},
},
},
},
{
name: "incoming pod's anti-affinity has more than one term - existing pod violates partial term - case 2",
existingPods: []*v1.Pod{normalPodB},
nodes: []*v1.Node{nodeA},
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "affi-antiaffi"},
Spec: v1.PodSpec{
Affinity: &v1.Affinity{
PodAffinity: &v1.PodAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa", "bbb"),
},
PodAntiAffinity: &v1.PodAntiAffinity{
RequiredDuringSchedulingIgnoredDuringExecution: newPodAffinityTerms("aaa", "bbb"),
},
},
},
},
wantAffinityPodsMaps: newTopologyPairsMaps(),
wantAntiAffinityPodsMaps: &topologyPairsMaps{
topologyPairToPods: map[topologyPair]podSet{
{key: "hostname", value: "nodeA"}: {normalPodB: struct{}{}},
},
podToTopologyPairs: map[string]topologyPairSet{
"normal_": {
topologyPair{key: "hostname", value: "nodeA"}: struct{}{},
},
},
},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
nodeInfoMap := schedulercache.CreateNodeNameToInfoMap(tt.existingPods, tt.nodes)
gotAffinityPodsMaps, gotAntiAffinityPodsMaps, err := getTPMapMatchingIncomingAffinityAntiAffinity(tt.pod, nodeInfoMap)
if (err != nil) != tt.wantErr {
t.Errorf("getTPMapMatchingIncomingAffinityAntiAffinity() error = %v, wantErr %v", err, tt.wantErr)
return
}
if !reflect.DeepEqual(gotAffinityPodsMaps, tt.wantAffinityPodsMaps) {
t.Errorf("getTPMapMatchingIncomingAffinityAntiAffinity() gotAffinityPodsMaps = %#v, want %#v", gotAffinityPodsMaps, tt.wantAffinityPodsMaps)
}
if !reflect.DeepEqual(gotAntiAffinityPodsMaps, tt.wantAntiAffinityPodsMaps) {
t.Errorf("getTPMapMatchingIncomingAffinityAntiAffinity() gotAntiAffinityPodsMaps = %#v, want %#v", gotAntiAffinityPodsMaps, tt.wantAntiAffinityPodsMaps)
}
})
}
}

View File

@ -20,10 +20,10 @@ import (
"errors"
"fmt"
"os"
"regexp"
"strconv"
"sync"
"github.com/golang/glog"
"k8s.io/klog"
"k8s.io/api/core/v1"
storagev1 "k8s.io/api/storage/v1"
@ -36,13 +36,13 @@ import (
utilfeature "k8s.io/apiserver/pkg/util/feature"
corelisters "k8s.io/client-go/listers/core/v1"
storagelisters "k8s.io/client-go/listers/storage/v1"
"k8s.io/client-go/util/workqueue"
v1helper "k8s.io/kubernetes/pkg/apis/core/v1/helper"
v1qos "k8s.io/kubernetes/pkg/apis/core/v1/helper/qos"
"k8s.io/kubernetes/pkg/features"
kubeletapis "k8s.io/kubernetes/pkg/kubelet/apis"
"k8s.io/kubernetes/pkg/scheduler/algorithm"
priorityutil "k8s.io/kubernetes/pkg/scheduler/algorithm/priorities/util"
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
schedutil "k8s.io/kubernetes/pkg/scheduler/util"
"k8s.io/kubernetes/pkg/scheduler/volumebinder"
@ -84,6 +84,8 @@ const (
MaxGCEPDVolumeCountPred = "MaxGCEPDVolumeCount"
// MaxAzureDiskVolumeCountPred defines the name of predicate MaxAzureDiskVolumeCount.
MaxAzureDiskVolumeCountPred = "MaxAzureDiskVolumeCount"
// MaxCSIVolumeCountPred defines the predicate that decides how many CSI volumes should be attached
MaxCSIVolumeCountPred = "MaxCSIVolumeCountPred"
// NoVolumeZoneConflictPred defines the name of predicate NoVolumeZoneConflict.
NoVolumeZoneConflictPred = "NoVolumeZoneConflict"
// CheckNodeMemoryPressurePred defines the name of predicate CheckNodeMemoryPressure.
@ -93,10 +95,6 @@ const (
// CheckNodePIDPressurePred defines the name of predicate CheckNodePIDPressure.
CheckNodePIDPressurePred = "CheckNodePIDPressure"
// DefaultMaxEBSVolumes is the limit for volumes attached to an instance.
// Amazon recommends no more than 40; the system root volume uses at least one.
// See http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/volume_limits.html#linux-specific-volume-limits
DefaultMaxEBSVolumes = 39
// DefaultMaxGCEPDVolumes defines the maximum number of PD Volumes for GCE
// GCE instances can have up to 16 PD volumes attached.
DefaultMaxGCEPDVolumes = 16
@ -134,7 +132,7 @@ var (
GeneralPred, HostNamePred, PodFitsHostPortsPred,
MatchNodeSelectorPred, PodFitsResourcesPred, NoDiskConflictPred,
PodToleratesNodeTaintsPred, PodToleratesNodeNoExecuteTaintsPred, CheckNodeLabelPresencePred,
CheckServiceAffinityPred, MaxEBSVolumeCountPred, MaxGCEPDVolumeCountPred,
CheckServiceAffinityPred, MaxEBSVolumeCountPred, MaxGCEPDVolumeCountPred, MaxCSIVolumeCountPred,
MaxAzureDiskVolumeCountPred, CheckVolumeBindingPred, NoVolumeZoneConflictPred,
CheckNodeMemoryPressurePred, CheckNodePIDPressurePred, CheckNodeDiskPressurePred, MatchInterPodAffinityPred}
)
@ -291,7 +289,7 @@ func NoDiskConflict(pod *v1.Pod, meta algorithm.PredicateMetadata, nodeInfo *sch
type MaxPDVolumeCountChecker struct {
filter VolumeFilter
volumeLimitKey v1.ResourceName
maxVolumes int
maxVolumeFunc func(node *v1.Node) int
pvInfo PersistentVolumeInfo
pvcInfo PersistentVolumeClaimInfo
@ -317,7 +315,6 @@ type VolumeFilter struct {
func NewMaxPDVolumeCountPredicate(
filterName string, pvInfo PersistentVolumeInfo, pvcInfo PersistentVolumeClaimInfo) algorithm.FitPredicate {
var filter VolumeFilter
var maxVolumes int
var volumeLimitKey v1.ResourceName
switch filterName {
@ -325,17 +322,14 @@ func NewMaxPDVolumeCountPredicate(
case EBSVolumeFilterType:
filter = EBSVolumeFilter
volumeLimitKey = v1.ResourceName(volumeutil.EBSVolumeLimitKey)
maxVolumes = getMaxVols(DefaultMaxEBSVolumes)
case GCEPDVolumeFilterType:
filter = GCEPDVolumeFilter
volumeLimitKey = v1.ResourceName(volumeutil.GCEVolumeLimitKey)
maxVolumes = getMaxVols(DefaultMaxGCEPDVolumes)
case AzureDiskVolumeFilterType:
filter = AzureDiskVolumeFilter
volumeLimitKey = v1.ResourceName(volumeutil.AzureVolumeLimitKey)
maxVolumes = getMaxVols(DefaultMaxAzureDiskVolumes)
default:
glog.Fatalf("Wrong filterName, Only Support %v %v %v ", EBSVolumeFilterType,
klog.Fatalf("Wrong filterName, Only Support %v %v %v ", EBSVolumeFilterType,
GCEPDVolumeFilterType, AzureDiskVolumeFilterType)
return nil
@ -343,7 +337,7 @@ func NewMaxPDVolumeCountPredicate(
c := &MaxPDVolumeCountChecker{
filter: filter,
volumeLimitKey: volumeLimitKey,
maxVolumes: maxVolumes,
maxVolumeFunc: getMaxVolumeFunc(filterName),
pvInfo: pvInfo,
pvcInfo: pvcInfo,
randomVolumeIDPrefix: rand.String(32),
@ -352,19 +346,52 @@ func NewMaxPDVolumeCountPredicate(
return c.predicate
}
// getMaxVols checks the max PD volumes environment variable, otherwise returning a default value
func getMaxVols(defaultVal int) int {
func getMaxVolumeFunc(filterName string) func(node *v1.Node) int {
return func(node *v1.Node) int {
maxVolumesFromEnv := getMaxVolLimitFromEnv()
if maxVolumesFromEnv > 0 {
return maxVolumesFromEnv
}
var nodeInstanceType string
for k, v := range node.ObjectMeta.Labels {
if k == kubeletapis.LabelInstanceType {
nodeInstanceType = v
}
}
switch filterName {
case EBSVolumeFilterType:
return getMaxEBSVolume(nodeInstanceType)
case GCEPDVolumeFilterType:
return DefaultMaxGCEPDVolumes
case AzureDiskVolumeFilterType:
return DefaultMaxAzureDiskVolumes
default:
return -1
}
}
}
func getMaxEBSVolume(nodeInstanceType string) int {
if ok, _ := regexp.MatchString(volumeutil.EBSNitroLimitRegex, nodeInstanceType); ok {
return volumeutil.DefaultMaxEBSNitroVolumeLimit
}
return volumeutil.DefaultMaxEBSVolumes
}
// getMaxVolLimitFromEnv checks the max PD volumes environment variable, otherwise returning a default value
func getMaxVolLimitFromEnv() int {
if rawMaxVols := os.Getenv(KubeMaxPDVols); rawMaxVols != "" {
if parsedMaxVols, err := strconv.Atoi(rawMaxVols); err != nil {
glog.Errorf("Unable to parse maximum PD volumes value, using default of %v: %v", defaultVal, err)
klog.Errorf("Unable to parse maximum PD volumes value, using default: %v", err)
} else if parsedMaxVols <= 0 {
glog.Errorf("Maximum PD volumes must be a positive value, using default of %v", defaultVal)
klog.Errorf("Maximum PD volumes must be a positive value, using default ")
} else {
return parsedMaxVols
}
}
return defaultVal
return -1
}
func (c *MaxPDVolumeCountChecker) filterVolumes(volumes []v1.Volume, namespace string, filteredVolumes map[string]bool) error {
@ -386,7 +413,7 @@ func (c *MaxPDVolumeCountChecker) filterVolumes(volumes []v1.Volume, namespace s
pvc, err := c.pvcInfo.GetPersistentVolumeClaimInfo(namespace, pvcName)
if err != nil || pvc == nil {
// if the PVC is not found, log the error and count the PV towards the PV limit
glog.V(4).Infof("Unable to look up PVC info for %s/%s, assuming PVC matches predicate when counting limits: %v", namespace, pvcName, err)
klog.V(4).Infof("Unable to look up PVC info for %s/%s, assuming PVC matches predicate when counting limits: %v", namespace, pvcName, err)
filteredVolumes[pvID] = true
continue
}
@ -397,7 +424,7 @@ func (c *MaxPDVolumeCountChecker) filterVolumes(volumes []v1.Volume, namespace s
// it was forcefully unbound by admin. The pod can still use the
// original PV where it was bound to -> log the error and count
// the PV towards the PV limit
glog.V(4).Infof("PVC %s/%s is not bound, assuming PVC matches predicate when counting limits", namespace, pvcName)
klog.V(4).Infof("PVC %s/%s is not bound, assuming PVC matches predicate when counting limits", namespace, pvcName)
filteredVolumes[pvID] = true
continue
}
@ -406,7 +433,7 @@ func (c *MaxPDVolumeCountChecker) filterVolumes(volumes []v1.Volume, namespace s
if err != nil || pv == nil {
// if the PV is not found, log the error
// and count the PV towards the PV limit
glog.V(4).Infof("Unable to look up PV info for %s/%s/%s, assuming PV matches predicate when counting limits: %v", namespace, pvcName, pvName, err)
klog.V(4).Infof("Unable to look up PV info for %s/%s/%s, assuming PV matches predicate when counting limits: %v", namespace, pvcName, pvName, err)
filteredVolumes[pvID] = true
continue
}
@ -454,7 +481,7 @@ func (c *MaxPDVolumeCountChecker) predicate(pod *v1.Pod, meta algorithm.Predicat
}
numNewVolumes := len(newVolumes)
maxAttachLimit := c.maxVolumes
maxAttachLimit := c.maxVolumeFunc(nodeInfo.Node())
if utilfeature.DefaultFeatureGate.Enabled(features.AttachVolumeLimit) {
volumeLimits := nodeInfo.VolumeLimits()
@ -638,12 +665,12 @@ func (c *VolumeZoneChecker) predicate(pod *v1.Pod, meta algorithm.PredicateMetad
nodeV, _ := nodeConstraints[k]
volumeVSet, err := volumeutil.LabelZonesToSet(v)
if err != nil {
glog.Warningf("Failed to parse label for %q: %q. Ignoring the label. err=%v. ", k, v, err)
klog.Warningf("Failed to parse label for %q: %q. Ignoring the label. err=%v. ", k, v, err)
continue
}
if !volumeVSet.Has(nodeV) {
glog.V(10).Infof("Won't schedule pod %q onto node %q due to volume %q (mismatch on %q)", pod.Name, node.Name, pvName, k)
klog.V(10).Infof("Won't schedule pod %q onto node %q due to volume %q (mismatch on %q)", pod.Name, node.Name, pvName, k)
return false, []algorithm.PredicateFailureReason{ErrVolumeZoneConflict}, nil
}
}
@ -754,11 +781,11 @@ func PodFitsResources(pod *v1.Pod, meta algorithm.PredicateMetadata, nodeInfo *s
}
}
if glog.V(10) {
if klog.V(10) {
if len(predicateFails) == 0 {
// We explicitly don't do glog.V(10).Infof() to avoid computing all the parameters if this is
// We explicitly don't do klog.V(10).Infof() to avoid computing all the parameters if this is
// not logged. There is visible performance gain from it.
glog.Infof("Schedule Pod %+v on Node %+v is allowed, Node is running only %v out of %v Pods.",
klog.Infof("Schedule Pod %+v on Node %+v is allowed, Node is running only %v out of %v Pods.",
podName(pod), node.Name, len(nodeInfo.Pods()), allowedPodNumber)
}
}
@ -807,14 +834,14 @@ func podMatchesNodeSelectorAndAffinityTerms(pod *v1.Pod, node *v1.Node) bool {
// TODO: Uncomment this block when implement RequiredDuringSchedulingRequiredDuringExecution.
// if nodeAffinity.RequiredDuringSchedulingRequiredDuringExecution != nil {
// nodeSelectorTerms := nodeAffinity.RequiredDuringSchedulingRequiredDuringExecution.NodeSelectorTerms
// glog.V(10).Infof("Match for RequiredDuringSchedulingRequiredDuringExecution node selector terms %+v", nodeSelectorTerms)
// klog.V(10).Infof("Match for RequiredDuringSchedulingRequiredDuringExecution node selector terms %+v", nodeSelectorTerms)
// nodeAffinityMatches = nodeMatchesNodeSelectorTerms(node, nodeSelectorTerms)
// }
// Match node selector for requiredDuringSchedulingIgnoredDuringExecution.
if nodeAffinity.RequiredDuringSchedulingIgnoredDuringExecution != nil {
nodeSelectorTerms := nodeAffinity.RequiredDuringSchedulingIgnoredDuringExecution.NodeSelectorTerms
glog.V(10).Infof("Match for RequiredDuringSchedulingIgnoredDuringExecution node selector terms %+v", nodeSelectorTerms)
klog.V(10).Infof("Match for RequiredDuringSchedulingIgnoredDuringExecution node selector terms %+v", nodeSelectorTerms)
nodeAffinityMatches = nodeAffinityMatches && nodeMatchesNodeSelectorTerms(node, nodeSelectorTerms)
}
@ -906,7 +933,7 @@ type ServiceAffinity struct {
// only should be referenced by NewServiceAffinityPredicate.
func (s *ServiceAffinity) serviceAffinityMetadataProducer(pm *predicateMetadata) {
if pm.pod == nil {
glog.Errorf("Cannot precompute service affinity, a pod is required to calculate service affinity.")
klog.Errorf("Cannot precompute service affinity, a pod is required to calculate service affinity.")
return
}
pm.serviceAffinityInUse = true
@ -918,7 +945,7 @@ func (s *ServiceAffinity) serviceAffinityMetadataProducer(pm *predicateMetadata)
// In the future maybe we will return them as part of the function.
if errSvc != nil || errList != nil {
glog.Errorf("Some Error were found while precomputing svc affinity: \nservices:%v , \npods:%v", errSvc, errList)
klog.Errorf("Some Error were found while precomputing svc affinity: \nservices:%v , \npods:%v", errSvc, errList)
}
// consider only the pods that belong to the same namespace
pm.serviceAffinityMatchingPodList = FilterPodsByNamespace(allMatches, pm.pod.Namespace)
@ -1145,10 +1172,10 @@ func (c *PodAffinityChecker) InterPodAffinityMatches(pod *v1.Pod, meta algorithm
return false, failedPredicates, error
}
if glog.V(10) {
// We explicitly don't do glog.V(10).Infof() to avoid computing all the parameters if this is
if klog.V(10) {
// We explicitly don't do klog.V(10).Infof() to avoid computing all the parameters if this is
// not logged. There is visible performance gain from it.
glog.Infof("Schedule Pod %+v on Node %+v is allowed, pod (anti)affinity constraints satisfied",
klog.Infof("Schedule Pod %+v on Node %+v is allowed, pod (anti)affinity constraints satisfied",
podName(pod), node.Name)
}
return true, nil, nil
@ -1159,7 +1186,7 @@ func (c *PodAffinityChecker) InterPodAffinityMatches(pod *v1.Pod, meta algorithm
// targetPod matches all the terms and their topologies, 2) whether targetPod
// matches all the terms label selector and namespaces (AKA term properties),
// 3) any error.
func (c *PodAffinityChecker) podMatchesPodAffinityTerms(pod *v1.Pod, targetPod *v1.Pod, nodeInfo *schedulercache.NodeInfo, terms []v1.PodAffinityTerm) (bool, bool, error) {
func (c *PodAffinityChecker) podMatchesPodAffinityTerms(pod, targetPod *v1.Pod, nodeInfo *schedulercache.NodeInfo, terms []v1.PodAffinityTerm) (bool, bool, error) {
if len(terms) == 0 {
return false, false, fmt.Errorf("terms array is empty")
}
@ -1167,7 +1194,7 @@ func (c *PodAffinityChecker) podMatchesPodAffinityTerms(pod *v1.Pod, targetPod *
if err != nil {
return false, false, err
}
if !podMatchesAffinityTermProperties(targetPod, props) {
if !podMatchesAllAffinityTermProperties(targetPod, props) {
return false, false, nil
}
// Namespace and selector of the terms have matched. Now we check topology of the terms.
@ -1214,186 +1241,129 @@ func GetPodAntiAffinityTerms(podAntiAffinity *v1.PodAntiAffinity) (terms []v1.Po
return terms
}
func getMatchingAntiAffinityTerms(pod *v1.Pod, nodeInfoMap map[string]*schedulercache.NodeInfo) (map[string][]matchingPodAntiAffinityTerm, error) {
allNodeNames := make([]string, 0, len(nodeInfoMap))
for name := range nodeInfoMap {
allNodeNames = append(allNodeNames, name)
// getMatchingAntiAffinityTopologyPairs calculates the following for "existingPod" on given node:
// (1) Whether it has PodAntiAffinity
// (2) Whether ANY AffinityTerm matches the incoming pod
func getMatchingAntiAffinityTopologyPairsOfPod(newPod *v1.Pod, existingPod *v1.Pod, node *v1.Node) (*topologyPairsMaps, error) {
affinity := existingPod.Spec.Affinity
if affinity == nil || affinity.PodAntiAffinity == nil {
return nil, nil
}
var lock sync.Mutex
var firstError error
result := make(map[string][]matchingPodAntiAffinityTerm)
appendResult := func(toAppend map[string][]matchingPodAntiAffinityTerm) {
lock.Lock()
defer lock.Unlock()
for uid, terms := range toAppend {
result[uid] = append(result[uid], terms...)
}
}
catchError := func(err error) {
lock.Lock()
defer lock.Unlock()
if firstError == nil {
firstError = err
topologyMaps := newTopologyPairsMaps()
for _, term := range GetPodAntiAffinityTerms(affinity.PodAntiAffinity) {
namespaces := priorityutil.GetNamespacesFromPodAffinityTerm(existingPod, &term)
selector, err := metav1.LabelSelectorAsSelector(term.LabelSelector)
if err != nil {
return nil, err
}
if priorityutil.PodMatchesTermsNamespaceAndSelector(newPod, namespaces, selector) {
if topologyValue, ok := node.Labels[term.TopologyKey]; ok {
pair := topologyPair{key: term.TopologyKey, value: topologyValue}
topologyMaps.addTopologyPair(pair, existingPod)
}
}
}
return topologyMaps, nil
}
processNode := func(i int) {
nodeInfo := nodeInfoMap[allNodeNames[i]]
node := nodeInfo.Node()
if node == nil {
catchError(fmt.Errorf("node not found"))
return
}
nodeResult := make(map[string][]matchingPodAntiAffinityTerm)
for _, existingPod := range nodeInfo.PodsWithAffinity() {
affinity := existingPod.Spec.Affinity
if affinity == nil {
func (c *PodAffinityChecker) getMatchingAntiAffinityTopologyPairsOfPods(pod *v1.Pod, existingPods []*v1.Pod) (*topologyPairsMaps, error) {
topologyMaps := newTopologyPairsMaps()
for _, existingPod := range existingPods {
existingPodNode, err := c.info.GetNodeInfo(existingPod.Spec.NodeName)
if err != nil {
if apierrors.IsNotFound(err) {
klog.Errorf("Node not found, %v", existingPod.Spec.NodeName)
continue
}
for _, term := range GetPodAntiAffinityTerms(affinity.PodAntiAffinity) {
namespaces := priorityutil.GetNamespacesFromPodAffinityTerm(existingPod, &term)
selector, err := metav1.LabelSelectorAsSelector(term.LabelSelector)
if err != nil {
catchError(err)
return
}
if priorityutil.PodMatchesTermsNamespaceAndSelector(pod, namespaces, selector) {
existingPodFullName := schedutil.GetPodFullName(existingPod)
nodeResult[existingPodFullName] = append(
nodeResult[existingPodFullName],
matchingPodAntiAffinityTerm{term: &term, node: node})
}
}
return nil, err
}
if len(nodeResult) > 0 {
appendResult(nodeResult)
existingPodTopologyMaps, err := getMatchingAntiAffinityTopologyPairsOfPod(pod, existingPod, existingPodNode)
if err != nil {
return nil, err
}
topologyMaps.appendMaps(existingPodTopologyMaps)
}
workqueue.Parallelize(16, len(allNodeNames), processNode)
return result, firstError
}
func getMatchingAntiAffinityTermsOfExistingPod(newPod *v1.Pod, existingPod *v1.Pod, node *v1.Node) ([]matchingPodAntiAffinityTerm, error) {
var result []matchingPodAntiAffinityTerm
affinity := existingPod.Spec.Affinity
if affinity != nil && affinity.PodAntiAffinity != nil {
for _, term := range GetPodAntiAffinityTerms(affinity.PodAntiAffinity) {
namespaces := priorityutil.GetNamespacesFromPodAffinityTerm(existingPod, &term)
selector, err := metav1.LabelSelectorAsSelector(term.LabelSelector)
if err != nil {
return nil, err
}
if priorityutil.PodMatchesTermsNamespaceAndSelector(newPod, namespaces, selector) {
result = append(result, matchingPodAntiAffinityTerm{term: &term, node: node})
}
}
}
return result, nil
}
func (c *PodAffinityChecker) getMatchingAntiAffinityTerms(pod *v1.Pod, allPods []*v1.Pod) (map[string][]matchingPodAntiAffinityTerm, error) {
result := make(map[string][]matchingPodAntiAffinityTerm)
for _, existingPod := range allPods {
affinity := existingPod.Spec.Affinity
if affinity != nil && affinity.PodAntiAffinity != nil {
existingPodNode, err := c.info.GetNodeInfo(existingPod.Spec.NodeName)
if err != nil {
if apierrors.IsNotFound(err) {
glog.Errorf("Node not found, %v", existingPod.Spec.NodeName)
continue
}
return nil, err
}
existingPodMatchingTerms, err := getMatchingAntiAffinityTermsOfExistingPod(pod, existingPod, existingPodNode)
if err != nil {
return nil, err
}
if len(existingPodMatchingTerms) > 0 {
existingPodFullName := schedutil.GetPodFullName(existingPod)
result[existingPodFullName] = existingPodMatchingTerms
}
}
}
return result, nil
return topologyMaps, nil
}
// Checks if scheduling the pod onto this node would break any anti-affinity
// rules indicated by the existing pods.
// terms indicated by the existing pods.
func (c *PodAffinityChecker) satisfiesExistingPodsAntiAffinity(pod *v1.Pod, meta algorithm.PredicateMetadata, nodeInfo *schedulercache.NodeInfo) (algorithm.PredicateFailureReason, error) {
node := nodeInfo.Node()
if node == nil {
return ErrExistingPodsAntiAffinityRulesNotMatch, fmt.Errorf("Node is nil")
}
var matchingTerms map[string][]matchingPodAntiAffinityTerm
var topologyMaps *topologyPairsMaps
if predicateMeta, ok := meta.(*predicateMetadata); ok {
matchingTerms = predicateMeta.matchingAntiAffinityTerms
topologyMaps = predicateMeta.topologyPairsAntiAffinityPodsMap
} else {
// Filter out pods whose nodeName is equal to nodeInfo.node.Name, but are not
// present in nodeInfo. Pods on other nodes pass the filter.
filteredPods, err := c.podLister.FilteredList(nodeInfo.Filter, labels.Everything())
if err != nil {
errMessage := fmt.Sprintf("Failed to get all pods, %+v", err)
glog.Error(errMessage)
klog.Error(errMessage)
return ErrExistingPodsAntiAffinityRulesNotMatch, errors.New(errMessage)
}
if matchingTerms, err = c.getMatchingAntiAffinityTerms(pod, filteredPods); err != nil {
if topologyMaps, err = c.getMatchingAntiAffinityTopologyPairsOfPods(pod, filteredPods); err != nil {
errMessage := fmt.Sprintf("Failed to get all terms that pod %+v matches, err: %+v", podName(pod), err)
glog.Error(errMessage)
klog.Error(errMessage)
return ErrExistingPodsAntiAffinityRulesNotMatch, errors.New(errMessage)
}
}
for _, terms := range matchingTerms {
for i := range terms {
term := &terms[i]
if len(term.term.TopologyKey) == 0 {
errMessage := fmt.Sprintf("Empty topologyKey is not allowed except for PreferredDuringScheduling pod anti-affinity")
glog.Error(errMessage)
return ErrExistingPodsAntiAffinityRulesNotMatch, errors.New(errMessage)
}
if priorityutil.NodesHaveSameTopologyKey(node, term.node, term.term.TopologyKey) {
glog.V(10).Infof("Cannot schedule pod %+v onto node %v,because of PodAntiAffinityTerm %v",
podName(pod), node.Name, term.term)
return ErrExistingPodsAntiAffinityRulesNotMatch, nil
}
// Iterate over topology pairs to get any of the pods being affected by
// the scheduled pod anti-affinity terms
for topologyKey, topologyValue := range node.Labels {
if topologyMaps.topologyPairToPods[topologyPair{key: topologyKey, value: topologyValue}] != nil {
klog.V(10).Infof("Cannot schedule pod %+v onto node %v", podName(pod), node.Name)
return ErrExistingPodsAntiAffinityRulesNotMatch, nil
}
}
if glog.V(10) {
// We explicitly don't do glog.V(10).Infof() to avoid computing all the parameters if this is
if klog.V(10) {
// We explicitly don't do klog.V(10).Infof() to avoid computing all the parameters if this is
// not logged. There is visible performance gain from it.
glog.Infof("Schedule Pod %+v on Node %+v is allowed, existing pods anti-affinity rules satisfied.",
klog.Infof("Schedule Pod %+v on Node %+v is allowed, existing pods anti-affinity terms satisfied.",
podName(pod), node.Name)
}
return nil, nil
}
// anyPodsMatchingTopologyTerms checks whether any of the nodes given via
// "targetPods" matches topology of all the "terms" for the give "pod" and "nodeInfo".
func (c *PodAffinityChecker) anyPodsMatchingTopologyTerms(pod *v1.Pod, targetPods map[string][]*v1.Pod, nodeInfo *schedulercache.NodeInfo, terms []v1.PodAffinityTerm) (bool, error) {
for nodeName, targetPods := range targetPods {
targetPodNodeInfo, err := c.info.GetNodeInfo(nodeName)
if err != nil {
return false, err
}
if len(targetPods) > 0 {
allTermsMatched := true
for _, term := range terms {
if !priorityutil.NodesHaveSameTopologyKey(nodeInfo.Node(), targetPodNodeInfo, term.TopologyKey) {
allTermsMatched = false
break
}
// nodeMatchesAllTopologyTerms checks whether "nodeInfo" matches
// topology of all the "terms" for the given "pod".
func (c *PodAffinityChecker) nodeMatchesAllTopologyTerms(pod *v1.Pod, topologyPairs *topologyPairsMaps, nodeInfo *schedulercache.NodeInfo, terms []v1.PodAffinityTerm) bool {
node := nodeInfo.Node()
for _, term := range terms {
if topologyValue, ok := node.Labels[term.TopologyKey]; ok {
pair := topologyPair{key: term.TopologyKey, value: topologyValue}
if _, ok := topologyPairs.topologyPairToPods[pair]; !ok {
return false
}
if allTermsMatched {
// We have 1 or more pods on the target node that have already matched namespace and selector
// and all of the terms topologies matched the target node. So, there is at least 1 matching pod on the node.
return true, nil
} else {
return false
}
}
return true
}
// nodeMatchesAnyTopologyTerm checks whether "nodeInfo" matches
// topology of any "term" for the given "pod".
func (c *PodAffinityChecker) nodeMatchesAnyTopologyTerm(pod *v1.Pod, topologyPairs *topologyPairsMaps, nodeInfo *schedulercache.NodeInfo, terms []v1.PodAffinityTerm) bool {
node := nodeInfo.Node()
for _, term := range terms {
if topologyValue, ok := node.Labels[term.TopologyKey]; ok {
pair := topologyPair{key: term.TopologyKey, value: topologyValue}
if _, ok := topologyPairs.topologyPairToPods[pair]; ok {
return true
}
}
}
return false, nil
return false
}
// Checks if scheduling the pod onto this node would break any rules of this pod.
// Checks if scheduling the pod onto this node would break any term of this pod.
func (c *PodAffinityChecker) satisfiesPodsAffinityAntiAffinity(pod *v1.Pod,
meta algorithm.PredicateMetadata, nodeInfo *schedulercache.NodeInfo,
affinity *v1.Affinity) (algorithm.PredicateFailureReason, error) {
@ -1403,21 +1373,16 @@ func (c *PodAffinityChecker) satisfiesPodsAffinityAntiAffinity(pod *v1.Pod,
}
if predicateMeta, ok := meta.(*predicateMetadata); ok {
// Check all affinity terms.
matchingPods := predicateMeta.nodeNameToMatchingAffinityPods
topologyPairsPotentialAffinityPods := predicateMeta.topologyPairsPotentialAffinityPods
if affinityTerms := GetPodAffinityTerms(affinity.PodAffinity); len(affinityTerms) > 0 {
matchExists, err := c.anyPodsMatchingTopologyTerms(pod, matchingPods, nodeInfo, affinityTerms)
if err != nil {
errMessage := fmt.Sprintf("Cannot schedule pod %+v onto node %v, because of PodAffinity, err: %v", podName(pod), node.Name, err)
glog.Errorf(errMessage)
return ErrPodAffinityRulesNotMatch, errors.New(errMessage)
}
matchExists := c.nodeMatchesAllTopologyTerms(pod, topologyPairsPotentialAffinityPods, nodeInfo, affinityTerms)
if !matchExists {
// This pod may the first pod in a series that have affinity to themselves. In order
// to not leave such pods in pending state forever, we check that if no other pod
// in the cluster matches the namespace and selector of this pod and the pod matches
// its own terms, then we allow the pod to pass the affinity check.
if !(len(matchingPods) == 0 && targetPodMatchesAffinityOfPod(pod, pod)) {
glog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAffinity",
if !(len(topologyPairsPotentialAffinityPods.topologyPairToPods) == 0 && targetPodMatchesAffinityOfPod(pod, pod)) {
klog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAffinity",
podName(pod), node.Name)
return ErrPodAffinityRulesNotMatch, nil
}
@ -1425,16 +1390,16 @@ func (c *PodAffinityChecker) satisfiesPodsAffinityAntiAffinity(pod *v1.Pod,
}
// Check all anti-affinity terms.
matchingPods = predicateMeta.nodeNameToMatchingAntiAffinityPods
topologyPairsPotentialAntiAffinityPods := predicateMeta.topologyPairsPotentialAntiAffinityPods
if antiAffinityTerms := GetPodAntiAffinityTerms(affinity.PodAntiAffinity); len(antiAffinityTerms) > 0 {
matchExists, err := c.anyPodsMatchingTopologyTerms(pod, matchingPods, nodeInfo, antiAffinityTerms)
if err != nil || matchExists {
glog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAntiAffinity, err: %v",
podName(pod), node.Name, err)
matchExists := c.nodeMatchesAnyTopologyTerm(pod, topologyPairsPotentialAntiAffinityPods, nodeInfo, antiAffinityTerms)
if matchExists {
klog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAntiAffinity",
podName(pod), node.Name)
return ErrPodAntiAffinityRulesNotMatch, nil
}
}
} else { // We don't have precomputed metadata. We have to follow a slow path to check affinity rules.
} else { // We don't have precomputed metadata. We have to follow a slow path to check affinity terms.
filteredPods, err := c.podLister.FilteredList(nodeInfo.Filter, labels.Everything())
if err != nil {
return ErrPodAffinityRulesNotMatch, err
@ -1449,7 +1414,7 @@ func (c *PodAffinityChecker) satisfiesPodsAffinityAntiAffinity(pod *v1.Pod,
affTermsMatch, termsSelectorMatch, err := c.podMatchesPodAffinityTerms(pod, targetPod, nodeInfo, affinityTerms)
if err != nil {
errMessage := fmt.Sprintf("Cannot schedule pod %+v onto node %v, because of PodAffinity, err: %v", podName(pod), node.Name, err)
glog.Error(errMessage)
klog.Error(errMessage)
return ErrPodAffinityRulesNotMatch, errors.New(errMessage)
}
if termsSelectorMatch {
@ -1464,7 +1429,7 @@ func (c *PodAffinityChecker) satisfiesPodsAffinityAntiAffinity(pod *v1.Pod,
if len(antiAffinityTerms) > 0 {
antiAffTermsMatch, _, err := c.podMatchesPodAffinityTerms(pod, targetPod, nodeInfo, antiAffinityTerms)
if err != nil || antiAffTermsMatch {
glog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAntiAffinityTerm, err: %v",
klog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAntiAffinityTerm, err: %v",
podName(pod), node.Name, err)
return ErrPodAntiAffinityRulesNotMatch, nil
}
@ -1472,29 +1437,29 @@ func (c *PodAffinityChecker) satisfiesPodsAffinityAntiAffinity(pod *v1.Pod,
}
if !matchFound && len(affinityTerms) > 0 {
// We have not been able to find any matches for the pod's affinity rules.
// We have not been able to find any matches for the pod's affinity terms.
// This pod may be the first pod in a series that have affinity to themselves. In order
// to not leave such pods in pending state forever, we check that if no other pod
// in the cluster matches the namespace and selector of this pod and the pod matches
// its own terms, then we allow the pod to pass the affinity check.
if termsSelectorMatchFound {
glog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAffinity",
klog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAffinity",
podName(pod), node.Name)
return ErrPodAffinityRulesNotMatch, nil
}
// Check if pod matches its own affinity properties (namespace and label selector).
if !targetPodMatchesAffinityOfPod(pod, pod) {
glog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAffinity",
klog.V(10).Infof("Cannot schedule pod %+v onto node %v, because of PodAffinity",
podName(pod), node.Name)
return ErrPodAffinityRulesNotMatch, nil
}
}
}
if glog.V(10) {
// We explicitly don't do glog.V(10).Infof() to avoid computing all the parameters if this is
if klog.V(10) {
// We explicitly don't do klog.V(10).Infof() to avoid computing all the parameters if this is
// not logged. There is visible performance gain from it.
glog.Infof("Schedule Pod %+v on Node %+v is allowed, pod affinity/anti-affinity constraints satisfied.",
klog.Infof("Schedule Pod %+v on Node %+v is allowed, pod affinity/anti-affinity constraints satisfied.",
podName(pod), node.Name)
}
return nil, nil
@ -1506,7 +1471,14 @@ func CheckNodeUnschedulablePredicate(pod *v1.Pod, meta algorithm.PredicateMetada
return false, []algorithm.PredicateFailureReason{ErrNodeUnknownCondition}, nil
}
if nodeInfo.Node().Spec.Unschedulable {
// If pod tolerate unschedulable taint, it's also tolerate `node.Spec.Unschedulable`.
podToleratesUnschedulable := v1helper.TolerationsTolerateTaint(pod.Spec.Tolerations, &v1.Taint{
Key: schedulerapi.TaintNodeUnschedulable,
Effect: v1.TaintEffectNoSchedule,
})
// TODO (k82cn): deprecates `node.Spec.Unschedulable` in 1.13.
if nodeInfo.Node().Spec.Unschedulable && !podToleratesUnschedulable {
return false, []algorithm.PredicateFailureReason{ErrNodeUnschedulable}, nil
}
@ -1662,12 +1634,12 @@ func (c *VolumeBindingChecker) predicate(pod *v1.Pod, meta algorithm.PredicateMe
failReasons := []algorithm.PredicateFailureReason{}
if !boundSatisfied {
glog.V(5).Infof("Bound PVs not satisfied for pod %v/%v, node %q", pod.Namespace, pod.Name, node.Name)
klog.V(5).Infof("Bound PVs not satisfied for pod %v/%v, node %q", pod.Namespace, pod.Name, node.Name)
failReasons = append(failReasons, ErrVolumeNodeConflict)
}
if !unboundSatisfied {
glog.V(5).Infof("Couldn't find matching PVs for pod %v/%v, node %q", pod.Namespace, pod.Name, node.Name)
klog.V(5).Infof("Couldn't find matching PVs for pod %v/%v, node %q", pod.Namespace, pod.Name, node.Name)
failReasons = append(failReasons, ErrVolumeBindConflict)
}
@ -1676,6 +1648,6 @@ func (c *VolumeBindingChecker) predicate(pod *v1.Pod, meta algorithm.PredicateMe
}
// All volumes bound or matching PVs found for all unbound PVCs
glog.V(5).Infof("All PVCs found matches for pod %v/%v, node %q", pod.Namespace, pod.Name, node.Name)
klog.V(5).Infof("All PVCs found matches for pod %v/%v, node %q", pod.Namespace, pod.Name, node.Name)
return true, nil, nil
}

File diff suppressed because it is too large Load Diff

View File

@ -19,7 +19,7 @@ package predicates
import (
"k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/labels"
schedutil "k8s.io/kubernetes/pkg/scheduler/util"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
)
// FindLabelsInSet gets as many key/value pairs as possible out of a label set.
@ -68,7 +68,7 @@ func CreateSelectorFromLabels(aL map[string]string) labels.Selector {
// portsConflict check whether existingPorts and wantPorts conflict with each other
// return true if we have a conflict
func portsConflict(existingPorts schedutil.HostPortInfo, wantPorts []*v1.ContainerPort) bool {
func portsConflict(existingPorts schedulercache.HostPortInfo, wantPorts []*v1.ContainerPort) bool {
for _, cp := range wantPorts {
if existingPorts.CheckConflict(cp.HostIP, string(cp.Protocol), cp.HostPort) {
return true

View File

@ -37,14 +37,14 @@ go_library(
"//pkg/scheduler/cache:go_default_library",
"//pkg/util/node:go_default_library",
"//pkg/util/parsers:go_default_library",
"//vendor/github.com/golang/glog:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/errors:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//vendor/k8s.io/client-go/util/workqueue:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/errors:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/client-go/util/workqueue:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
@ -55,6 +55,7 @@ go_test(
"image_locality_test.go",
"interpod_affinity_test.go",
"least_requested_test.go",
"main_test.go",
"metadata_test.go",
"most_requested_test.go",
"node_affinity_test.go",
@ -74,13 +75,13 @@ go_test(
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/testing:go_default_library",
"//pkg/util/parsers:go_default_library",
"//staging/src/k8s.io/api/apps/v1:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature/testing:go_default_library",
"//vendor/github.com/stretchr/testify/assert:go_default_library",
"//vendor/k8s.io/api/apps/v1beta1:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/extensions/v1beta1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
],
)

View File

@ -32,7 +32,7 @@ var (
// BalancedResourceAllocationMap should **NOT** be used alone, and **MUST** be used together
// with LeastRequestedPriority. It calculates the difference between the cpu and memory fraction
// 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:
// Detail: score = 10 - variance(cpuFraction,memoryFraction,volumeFraction)*10. The algorithm is partly inspired by:
// "Wei Huang et al. An Energy Efficient Virtual Machine Placement Algorithm with Balanced
// Resource Utilization"
BalancedResourceAllocationMap = balancedResourcePriority.PriorityMap

View File

@ -17,7 +17,6 @@ limitations under the License.
package priorities
import (
"fmt"
"reflect"
"testing"
@ -25,6 +24,7 @@ import (
"k8s.io/apimachinery/pkg/api/resource"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
utilfeature "k8s.io/apiserver/pkg/util/feature"
utilfeaturetesting "k8s.io/apiserver/pkg/util/feature/testing"
"k8s.io/kubernetes/pkg/features"
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
@ -44,7 +44,7 @@ func getExistingVolumeCountForNode(pods []*v1.Pod, maxVolumes int) int {
func TestBalancedResourceAllocation(t *testing.T) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
podwithVol1 := v1.PodSpec{
Containers: []v1.Container{
{

View File

@ -26,11 +26,12 @@ import (
"k8s.io/kubernetes/pkg/util/parsers"
)
// This is a reasonable size range of all container images. 90%ile of images on dockerhub drops into this range.
// The two thresholds are used as bounds for the image score range. They correspond to a reasonable size range for
// container images compressed and stored in registries; 90%ile of images on dockerhub drops into this range.
const (
mb int64 = 1024 * 1024
minImgSize int64 = 23 * mb
maxImgSize int64 = 1000 * mb
mb int64 = 1024 * 1024
minThreshold int64 = 23 * mb
maxThreshold int64 = 1000 * mb
)
// ImageLocalityPriorityMap is a priority function that favors nodes that already have requested pod container's images.
@ -44,44 +45,55 @@ func ImageLocalityPriorityMap(pod *v1.Pod, meta interface{}, nodeInfo *scheduler
return schedulerapi.HostPriority{}, fmt.Errorf("node not found")
}
sumSize := totalImageSize(nodeInfo, pod.Spec.Containers)
var score int
if priorityMeta, ok := meta.(*priorityMetadata); ok {
score = calculatePriority(sumImageScores(nodeInfo, pod.Spec.Containers, priorityMeta.totalNumNodes))
} else {
// if we are not able to parse priority meta data, skip this priority
score = 0
}
return schedulerapi.HostPriority{
Host: node.Name,
Score: calculateScoreFromSize(sumSize),
Score: score,
}, nil
}
// calculateScoreFromSize calculates the priority of a node. sumSize is sum size of requested images on this node.
// 1. Split image size range into 10 buckets.
// 2. Decide the priority of a given sumSize based on which bucket it belongs to.
func calculateScoreFromSize(sumSize int64) int {
switch {
case sumSize == 0 || sumSize < minImgSize:
// 0 means none of the images required by this pod are present on this
// node or the total size of the images present is too small to be taken into further consideration.
return 0
case sumSize >= maxImgSize:
// If existing images' total size is larger than max, just make it highest priority.
return schedulerapi.MaxPriority
// calculatePriority returns the priority of a node. Given the sumScores of requested images on the node, the node's
// priority is obtained by scaling the maximum priority value with a ratio proportional to the sumScores.
func calculatePriority(sumScores int64) int {
if sumScores < minThreshold {
sumScores = minThreshold
} else if sumScores > maxThreshold {
sumScores = maxThreshold
}
return int((int64(schedulerapi.MaxPriority) * (sumSize - minImgSize) / (maxImgSize - minImgSize)) + 1)
return int(int64(schedulerapi.MaxPriority) * (sumScores - minThreshold) / (maxThreshold - minThreshold))
}
// totalImageSize returns the total image size of all the containers that are already on the node.
func totalImageSize(nodeInfo *schedulercache.NodeInfo, containers []v1.Container) int64 {
var total int64
// sumImageScores returns the sum of image scores of all the containers that are already on the node.
// Each image receives a raw score of its size, scaled by scaledImageScore. The raw scores are later used to calculate
// the final score. Note that the init containers are not considered for it's rare for users to deploy huge init containers.
func sumImageScores(nodeInfo *schedulercache.NodeInfo, containers []v1.Container, totalNumNodes int) int64 {
var sum int64
imageStates := nodeInfo.ImageStates()
imageSizes := nodeInfo.ImageSizes()
for _, container := range containers {
if size, ok := imageSizes[normalizedImageName(container.Image)]; ok {
total += size
if state, ok := imageStates[normalizedImageName(container.Image)]; ok {
sum += scaledImageScore(state, totalNumNodes)
}
}
return total
return sum
}
// scaledImageScore returns an adaptively scaled score for the given state of an image.
// The size of the image is used as the base score, scaled by a factor which considers how much nodes the image has "spread" to.
// This heuristic aims to mitigate the undesirable "node heating problem", i.e., pods get assigned to the same or
// a few nodes due to image locality.
func scaledImageScore(imageState *schedulercache.ImageStateSummary, totalNumNodes int) int64 {
spread := float64(imageState.NumNodes) / float64(totalNumNodes)
return int64(float64(imageState.Size) * spread)
}
// normalizedImageName returns the CRI compliant name for a given image.

View File

@ -42,13 +42,13 @@ func TestImageLocalityPriority(t *testing.T) {
},
}
test40140 := v1.PodSpec{
test40300 := v1.PodSpec{
Containers: []v1.Container{
{
Image: "gcr.io/40",
},
{
Image: "gcr.io/140",
Image: "gcr.io/300",
},
},
}
@ -64,7 +64,7 @@ func TestImageLocalityPriority(t *testing.T) {
},
}
node401402000 := v1.NodeStatus{
node403002000 := v1.NodeStatus{
Images: []v1.ContainerImage{
{
Names: []string{
@ -76,10 +76,10 @@ func TestImageLocalityPriority(t *testing.T) {
},
{
Names: []string{
"gcr.io/140:" + parsers.DefaultImageTag,
"gcr.io/140:v1",
"gcr.io/300:" + parsers.DefaultImageTag,
"gcr.io/300:v1",
},
SizeBytes: int64(140 * mb),
SizeBytes: int64(300 * mb),
},
{
Names: []string{
@ -120,29 +120,29 @@ func TestImageLocalityPriority(t *testing.T) {
// Node1
// Image: gcr.io/40:latest 40MB
// Score: (40M-23M)/97.7M + 1 = 1
// Score: 0 (40M/2 < 23M, min-threshold)
// Node2
// Image: gcr.io/250:latest 250MB
// Score: (250M-23M)/97.7M + 1 = 3
// Score: 10 * (250M/2 - 23M)/(1000M - 23M) = 1
pod: &v1.Pod{Spec: test40250},
nodes: []*v1.Node{makeImageNode("machine1", node401402000), makeImageNode("machine2", node25010)},
expectedList: []schedulerapi.HostPriority{{Host: "machine1", Score: 1}, {Host: "machine2", Score: 3}},
nodes: []*v1.Node{makeImageNode("machine1", node403002000), makeImageNode("machine2", node25010)},
expectedList: []schedulerapi.HostPriority{{Host: "machine1", Score: 0}, {Host: "machine2", Score: 1}},
name: "two images spread on two nodes, prefer the larger image one",
},
{
// Pod: gcr.io/40 gcr.io/140
// Pod: gcr.io/40 gcr.io/300
// Node1
// Image: gcr.io/40:latest 40MB, gcr.io/140:latest 140MB
// Score: (40M+140M-23M)/97.7M + 1 = 2
// Image: gcr.io/40:latest 40MB, gcr.io/300:latest 300MB
// Score: 10 * ((40M + 300M)/2 - 23M)/(1000M - 23M) = 1
// Node2
// Image: not present
// Score: 0
pod: &v1.Pod{Spec: test40140},
nodes: []*v1.Node{makeImageNode("machine1", node401402000), makeImageNode("machine2", node25010)},
expectedList: []schedulerapi.HostPriority{{Host: "machine1", Score: 2}, {Host: "machine2", Score: 0}},
pod: &v1.Pod{Spec: test40300},
nodes: []*v1.Node{makeImageNode("machine1", node403002000), makeImageNode("machine2", node25010)},
expectedList: []schedulerapi.HostPriority{{Host: "machine1", Score: 1}, {Host: "machine2", Score: 0}},
name: "two images on one node, prefer this node",
},
{
@ -150,13 +150,13 @@ func TestImageLocalityPriority(t *testing.T) {
// Node1
// Image: gcr.io/2000:latest 2000MB
// Score: 2000 > max score = 10
// Score: 10 (2000M/2 >= 1000M, max-threshold)
// Node2
// Image: gcr.io/10:latest 10MB
// Score: 10 < min score = 0
// Score: 0 (10M/2 < 23M, min-threshold)
pod: &v1.Pod{Spec: testMinMax},
nodes: []*v1.Node{makeImageNode("machine1", node401402000), makeImageNode("machine2", node25010)},
nodes: []*v1.Node{makeImageNode("machine1", node403002000), makeImageNode("machine2", node25010)},
expectedList: []schedulerapi.HostPriority{{Host: "machine1", Score: schedulerapi.MaxPriority}, {Host: "machine2", Score: 0}},
name: "if exceed limit, use limit",
},
@ -165,7 +165,7 @@ func TestImageLocalityPriority(t *testing.T) {
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
nodeNameToInfo := schedulercache.CreateNodeNameToInfoMap(test.pods, test.nodes)
list, err := priorityFunction(ImageLocalityPriorityMap, nil, nil)(test.pod, nodeNameToInfo, test.nodes)
list, err := priorityFunction(ImageLocalityPriorityMap, nil, &priorityMetadata{totalNumNodes: len(test.nodes)})(test.pod, nodeNameToInfo, test.nodes)
if err != nil {
t.Errorf("unexpected error: %v", err)
}

View File

@ -17,6 +17,7 @@ limitations under the License.
package priorities
import (
"context"
"sync"
"k8s.io/api/core/v1"
@ -29,7 +30,7 @@ import (
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
"github.com/golang/glog"
"k8s.io/klog"
)
// InterPodAffinity contains information to calculate inter pod affinity.
@ -136,7 +137,7 @@ func (ipa *InterPodAffinity) CalculateInterPodAffinityPriority(pod *v1.Pod, node
existingPodNode, err := ipa.info.GetNodeInfo(existingPod.Spec.NodeName)
if err != nil {
if apierrors.IsNotFound(err) {
glog.Errorf("Node not found, %v", existingPod.Spec.NodeName)
klog.Errorf("Node not found, %v", existingPod.Spec.NodeName)
return nil
}
return err
@ -210,7 +211,7 @@ func (ipa *InterPodAffinity) CalculateInterPodAffinityPriority(pod *v1.Pod, node
}
}
}
workqueue.Parallelize(16, len(allNodeNames), processNode)
workqueue.ParallelizeUntil(context.TODO(), 16, len(allNodeNames), processNode)
if pm.firstError != nil {
return nil, pm.firstError
}
@ -232,8 +233,8 @@ func (ipa *InterPodAffinity) CalculateInterPodAffinityPriority(pod *v1.Pod, node
fScore = float64(schedulerapi.MaxPriority) * ((pm.counts[node.Name] - minCount) / (maxCount - minCount))
}
result = append(result, schedulerapi.HostPriority{Host: node.Name, Score: int(fScore)})
if glog.V(10) {
glog.Infof("%v -> %v: InterPodAffinityPriority, Score: (%d)", pod.Name, node.Name, int(fScore))
if klog.V(10) {
klog.Infof("%v -> %v: InterPodAffinityPriority, Score: (%d)", pod.Name, node.Name, int(fScore))
}
}
return result, nil

View File

@ -29,7 +29,7 @@ var (
// 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
// (cpu((capacity-sum(requested))*10/capacity) + memory((capacity-sum(requested))*10/capacity))/2
LeastRequestedPriorityMap = leastResourcePriority.PriorityMap
)

View File

@ -0,0 +1,29 @@
/*
Copyright 2018 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 priorities
import (
"testing"
utilfeature "k8s.io/apiserver/pkg/util/feature"
utilfeaturetesting "k8s.io/apiserver/pkg/util/feature/testing"
_ "k8s.io/kubernetes/pkg/features"
)
func TestMain(m *testing.M) {
utilfeaturetesting.VerifyFeatureGatesUnchanged(utilfeature.DefaultFeatureGate, m.Run)
}

View File

@ -21,7 +21,6 @@ import (
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/labels"
"k8s.io/kubernetes/pkg/scheduler/algorithm"
priorityutil "k8s.io/kubernetes/pkg/scheduler/algorithm/priorities/util"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
)
@ -52,6 +51,7 @@ type priorityMetadata struct {
podSelectors []labels.Selector
controllerRef *metav1.OwnerReference
podFirstServiceSelector labels.Selector
totalNumNodes int
}
// PriorityMetadata is a PriorityMetadataProducer. Node info can be nil.
@ -65,8 +65,9 @@ func (pmf *PriorityMetadataFactory) PriorityMetadata(pod *v1.Pod, nodeNameToInfo
podTolerations: getAllTolerationPreferNoSchedule(pod.Spec.Tolerations),
affinity: pod.Spec.Affinity,
podSelectors: getSelectors(pod, pmf.serviceLister, pmf.controllerLister, pmf.replicaSetLister, pmf.statefulSetLister),
controllerRef: priorityutil.GetControllerRef(pod),
controllerRef: metav1.GetControllerOf(pod),
podFirstServiceSelector: getFirstServiceSelector(pod, pmf.serviceLister),
totalNumNodes: len(nodeNameToInfo),
}
}

View File

@ -20,9 +20,8 @@ import (
"reflect"
"testing"
apps "k8s.io/api/apps/v1beta1"
apps "k8s.io/api/apps/v1"
"k8s.io/api/core/v1"
extensions "k8s.io/api/extensions/v1beta1"
"k8s.io/apimachinery/pkg/api/resource"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
priorityutil "k8s.io/kubernetes/pkg/scheduler/algorithm/priorities/util"
@ -153,14 +152,14 @@ func TestPriorityMetadata(t *testing.T) {
name: "Produce a priorityMetadata with specified requests",
},
}
mataDataProducer := NewPriorityMetadataFactory(
metaDataProducer := NewPriorityMetadataFactory(
schedulertesting.FakeServiceLister([]*v1.Service{}),
schedulertesting.FakeControllerLister([]*v1.ReplicationController{}),
schedulertesting.FakeReplicaSetLister([]*extensions.ReplicaSet{}),
schedulertesting.FakeReplicaSetLister([]*apps.ReplicaSet{}),
schedulertesting.FakeStatefulSetLister([]*apps.StatefulSet{}))
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
ptData := mataDataProducer(test.pod, nil)
ptData := metaDataProducer(test.pod, nil)
if !reflect.DeepEqual(test.expected, ptData) {
t.Errorf("expected %#v, got %#v", test.expected, ptData)
}

View File

@ -39,10 +39,10 @@ func mostResourceScorer(requested, allocable *schedulercache.Resource, includeVo
// The used capacity is calculated on a scale of 0-10
// 0 being the lowest priority and 10 being the highest.
// The more resources are used the higher the score is. This function
// is almost a reversed version of least_requested_priority.calculatUnusedScore
// is almost a reversed version of least_requested_priority.calculateUnusedScore
// (10 - calculateUnusedScore). The main difference is in rounding. It was added to
// keep the final formula clean and not to modify the widely used (by users
// in their default scheduling policies) calculateUSedScore.
// in their default scheduling policies) calculateUsedScore.
func mostRequestedScore(requested, capacity int64) int64 {
if capacity == 0 {
return 0

View File

@ -22,7 +22,6 @@ import (
"k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
v1helper "k8s.io/kubernetes/pkg/apis/core/v1/helper"
priorityutil "k8s.io/kubernetes/pkg/scheduler/algorithm/priorities/util"
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
)
@ -39,7 +38,7 @@ func CalculateNodePreferAvoidPodsPriorityMap(pod *v1.Pod, meta interface{}, node
controllerRef = priorityMeta.controllerRef
} else {
// We couldn't parse metadata - fallback to the podspec.
controllerRef = priorityutil.GetControllerRef(pod)
controllerRef = metav1.GetControllerOf(pod)
}
if controllerRef != nil {

View File

@ -209,19 +209,19 @@ func TestRequestedToCapacityRatio(t *testing.T) {
for _, test := range tests {
nodeNames := make([]string, 0)
var nodeNames []string
for nodeName := range test.nodes {
nodeNames = append(nodeNames, nodeName)
}
sort.Strings(nodeNames)
nodes := make([]*v1.Node, 0)
var nodes []*v1.Node
for _, nodeName := range nodeNames {
node := test.nodes[nodeName]
nodes = append(nodes, makeNode(nodeName, node.capacity.cpu, node.capacity.mem))
}
scheduledPods := make([]*v1.Pod, 0)
var scheduledPods []*v1.Pod
for name, node := range test.nodes {
scheduledPods = append(scheduledPods,
buildResourcesPod(name, node.used))

View File

@ -19,8 +19,10 @@ package priorities
import (
"fmt"
"github.com/golang/glog"
"k8s.io/api/core/v1"
utilfeature "k8s.io/apiserver/pkg/util/feature"
"k8s.io/klog"
"k8s.io/kubernetes/pkg/features"
priorityutil "k8s.io/kubernetes/pkg/scheduler/algorithm/priorities/util"
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
@ -56,20 +58,31 @@ func (r *ResourceAllocationPriority) PriorityMap(
requested.Memory += nodeInfo.NonZeroRequest().Memory
var score int64
// Check if the pod has volumes and this could be added to scorer function for balanced resource allocation.
if len(pod.Spec.Volumes) >= 0 && nodeInfo.TransientInfo != nil {
if len(pod.Spec.Volumes) >= 0 && utilfeature.DefaultFeatureGate.Enabled(features.BalanceAttachedNodeVolumes) && nodeInfo.TransientInfo != nil {
score = r.scorer(&requested, &allocatable, true, nodeInfo.TransientInfo.TransNodeInfo.RequestedVolumes, nodeInfo.TransientInfo.TransNodeInfo.AllocatableVolumesCount)
} else {
score = r.scorer(&requested, &allocatable, false, 0, 0)
}
if glog.V(10) {
glog.Infof(
"%v -> %v: %v, capacity %d millicores %d memory bytes, total request %d millicores %d memory bytes, score %d",
pod.Name, node.Name, r.Name,
allocatable.MilliCPU, allocatable.Memory,
requested.MilliCPU+allocatable.MilliCPU, requested.Memory+allocatable.Memory,
score,
)
if klog.V(10) {
if len(pod.Spec.Volumes) >= 0 && utilfeature.DefaultFeatureGate.Enabled(features.BalanceAttachedNodeVolumes) && nodeInfo.TransientInfo != nil {
klog.Infof(
"%v -> %v: %v, capacity %d millicores %d memory bytes, %d volumes, total request %d millicores %d memory bytes %d volumes, score %d",
pod.Name, node.Name, r.Name,
allocatable.MilliCPU, allocatable.Memory, nodeInfo.TransientInfo.TransNodeInfo.AllocatableVolumesCount,
requested.MilliCPU, requested.Memory,
nodeInfo.TransientInfo.TransNodeInfo.RequestedVolumes,
score,
)
} else {
klog.Infof(
"%v -> %v: %v, capacity %d millicores %d memory bytes, total request %d millicores %d memory bytes, score %d",
pod.Name, node.Name, r.Name,
allocatable.MilliCPU, allocatable.Memory,
requested.MilliCPU, requested.Memory,
score,
)
}
}
return schedulerapi.HostPriority{

View File

@ -23,7 +23,7 @@ import (
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
"github.com/golang/glog"
"k8s.io/klog"
)
// ResourceLimitsPriorityMap is a priority function that increases score of input node by 1 if the node satisfies
@ -52,10 +52,10 @@ func ResourceLimitsPriorityMap(pod *v1.Pod, meta interface{}, nodeInfo *schedule
score = 1
}
if glog.V(10) {
// We explicitly don't do glog.V(10).Infof() to avoid computing all the parameters if this is
if klog.V(10) {
// We explicitly don't do klog.V(10).Infof() to avoid computing all the parameters if this is
// not logged. There is visible performance gain from it.
glog.Infof(
klog.Infof(
"%v -> %v: Resource Limits Priority, allocatable %d millicores %d memory bytes, pod limits %d millicores %d memory bytes, score %d",
pod.Name, node.Name,
allocatableResources.MilliCPU, allocatableResources.Memory,
@ -70,7 +70,7 @@ func ResourceLimitsPriorityMap(pod *v1.Pod, meta interface{}, nodeInfo *schedule
}, nil
}
// computeScore return 1 if limit value is less than or equal to allocable
// computeScore returns 1 if limit value is less than or equal to allocatable
// value, otherwise it returns 0.
func computeScore(limit, allocatable int64) int64 {
if limit != 0 && allocatable != 0 && limit <= allocatable {

View File

@ -26,7 +26,7 @@ import (
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
utilnode "k8s.io/kubernetes/pkg/util/node"
"github.com/golang/glog"
"k8s.io/klog"
)
// When zone information is present, give 2/3 of the weighting to zone spreading, 1/3 to node spreading
@ -94,19 +94,15 @@ func (s *SelectorSpread) CalculateSpreadPriorityMap(pod *v1.Pod, meta interface{
// Ignore the previous deleted version for spreading purposes
// (it can still be considered for resource restrictions etc.)
if nodePod.DeletionTimestamp != nil {
glog.V(4).Infof("skipping pending-deleted pod: %s/%s", nodePod.Namespace, nodePod.Name)
klog.V(4).Infof("skipping pending-deleted pod: %s/%s", nodePod.Namespace, nodePod.Name)
continue
}
matches := false
for _, selector := range selectors {
if selector.Matches(labels.Set(nodePod.ObjectMeta.Labels)) {
matches = true
count++
break
}
}
if matches {
count++
}
}
return schedulerapi.HostPriority{
Host: node.Name,
@ -164,8 +160,8 @@ func (s *SelectorSpread) CalculateSpreadPriorityReduce(pod *v1.Pod, meta interfa
}
}
result[i].Score = int(fScore)
if glog.V(10) {
glog.Infof(
if klog.V(10) {
klog.Infof(
"%v -> %v: SelectorSpreadPriority, Score: (%d)", pod.Name, result[i].Host, int(fScore),
)
}

View File

@ -21,9 +21,8 @@ import (
"sort"
"testing"
apps "k8s.io/api/apps/v1beta1"
apps "k8s.io/api/apps/v1"
"k8s.io/api/core/v1"
extensions "k8s.io/api/extensions/v1beta1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
kubeletapis "k8s.io/kubernetes/pkg/kubelet/apis"
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
@ -60,7 +59,7 @@ func TestSelectorSpreadPriority(t *testing.T) {
pods []*v1.Pod
nodes []string
rcs []*v1.ReplicationController
rss []*extensions.ReplicaSet
rss []*apps.ReplicaSet
services []*v1.Service
sss []*apps.StatefulSet
expectedList schedulerapi.HostPriorityList
@ -200,7 +199,7 @@ func TestSelectorSpreadPriority(t *testing.T) {
},
nodes: []string{"machine1", "machine2"},
services: []*v1.Service{{Spec: v1.ServiceSpec{Selector: map[string]string{"baz": "blah"}}}},
rss: []*extensions.ReplicaSet{{Spec: extensions.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}}}}},
rss: []*apps.ReplicaSet{{Spec: apps.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}}}}},
// We use ReplicaSet, instead of ReplicationController. The result should be exactly as above.
expectedList: []schedulerapi.HostPriority{{Host: "machine1", Score: 0}, {Host: "machine2", Score: 5}},
name: "service with partial pod label matches with service and replica set",
@ -241,7 +240,7 @@ func TestSelectorSpreadPriority(t *testing.T) {
},
nodes: []string{"machine1", "machine2"},
services: []*v1.Service{{Spec: v1.ServiceSpec{Selector: map[string]string{"bar": "foo"}}}},
rss: []*extensions.ReplicaSet{{Spec: extensions.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}}}}},
rss: []*apps.ReplicaSet{{Spec: apps.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}}}}},
// We use ReplicaSet, instead of ReplicationController. The result should be exactly as above.
expectedList: []schedulerapi.HostPriority{{Host: "machine1", Score: 0}, {Host: "machine2", Score: 5}},
name: "disjoined service and replica set should be treated equally",
@ -280,7 +279,7 @@ func TestSelectorSpreadPriority(t *testing.T) {
{Spec: zone2Spec, ObjectMeta: metav1.ObjectMeta{Labels: labels1, OwnerReferences: controllerRef("ReplicaSet", "name", "abc123")}},
},
nodes: []string{"machine1", "machine2"},
rss: []*extensions.ReplicaSet{{Spec: extensions.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}}}}},
rss: []*apps.ReplicaSet{{Spec: apps.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}}}}},
// We use ReplicaSet, instead of ReplicationController. The result should be exactly as above.
expectedList: []schedulerapi.HostPriority{{Host: "machine1", Score: 0}, {Host: "machine2", Score: 0}},
name: "Replica set with partial pod label matches",
@ -318,7 +317,7 @@ func TestSelectorSpreadPriority(t *testing.T) {
{Spec: zone2Spec, ObjectMeta: metav1.ObjectMeta{Labels: labels1, OwnerReferences: controllerRef("ReplicaSet", "name", "abc123")}},
},
nodes: []string{"machine1", "machine2"},
rss: []*extensions.ReplicaSet{{Spec: extensions.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"baz": "blah"}}}}},
rss: []*apps.ReplicaSet{{Spec: apps.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"baz": "blah"}}}}},
// We use ReplicaSet, instead of ReplicationController. The result should be exactly as above.
expectedList: []schedulerapi.HostPriority{{Host: "machine1", Score: 0}, {Host: "machine2", Score: 5}},
name: "Another replication set with partial pod label matches",
@ -348,14 +347,14 @@ func TestSelectorSpreadPriority(t *testing.T) {
statefulSetLister: schedulertesting.FakeStatefulSetLister(test.sss),
}
mataDataProducer := NewPriorityMetadataFactory(
metaDataProducer := NewPriorityMetadataFactory(
schedulertesting.FakeServiceLister(test.services),
schedulertesting.FakeControllerLister(test.rcs),
schedulertesting.FakeReplicaSetLister(test.rss),
schedulertesting.FakeStatefulSetLister(test.sss))
mataData := mataDataProducer(test.pod, nodeNameToInfo)
metaData := metaDataProducer(test.pod, nodeNameToInfo)
ttp := priorityFunction(selectorSpread.CalculateSpreadPriorityMap, selectorSpread.CalculateSpreadPriorityReduce, mataData)
ttp := priorityFunction(selectorSpread.CalculateSpreadPriorityMap, selectorSpread.CalculateSpreadPriorityReduce, metaData)
list, err := ttp(test.pod, nodeNameToInfo, makeNodeList(test.nodes))
if err != nil {
t.Errorf("unexpected error: %v \n", err)
@ -410,7 +409,7 @@ func TestZoneSelectorSpreadPriority(t *testing.T) {
pod *v1.Pod
pods []*v1.Pod
rcs []*v1.ReplicationController
rss []*extensions.ReplicaSet
rss []*apps.ReplicaSet
services []*v1.Service
sss []*apps.StatefulSet
expectedList schedulerapi.HostPriorityList
@ -584,13 +583,13 @@ func TestZoneSelectorSpreadPriority(t *testing.T) {
statefulSetLister: schedulertesting.FakeStatefulSetLister(test.sss),
}
mataDataProducer := NewPriorityMetadataFactory(
metaDataProducer := NewPriorityMetadataFactory(
schedulertesting.FakeServiceLister(test.services),
schedulertesting.FakeControllerLister(test.rcs),
schedulertesting.FakeReplicaSetLister(test.rss),
schedulertesting.FakeStatefulSetLister(test.sss))
mataData := mataDataProducer(test.pod, nodeNameToInfo)
ttp := priorityFunction(selectorSpread.CalculateSpreadPriorityMap, selectorSpread.CalculateSpreadPriorityReduce, mataData)
metaData := metaDataProducer(test.pod, nodeNameToInfo)
ttp := priorityFunction(selectorSpread.CalculateSpreadPriorityMap, selectorSpread.CalculateSpreadPriorityReduce, metaData)
list, err := ttp(test.pod, nodeNameToInfo, makeLabeledNodeList(labeledNodes))
if err != nil {
t.Errorf("unexpected error: %v", err)
@ -761,23 +760,23 @@ func TestZoneSpreadPriority(t *testing.T) {
},
}
// these local variables just make sure controllerLister\replicaSetLister\statefulSetLister not nil
// when construct mataDataProducer
// when construct metaDataProducer
sss := []*apps.StatefulSet{{Spec: apps.StatefulSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}}}}}
rcs := []*v1.ReplicationController{{Spec: v1.ReplicationControllerSpec{Selector: map[string]string{"foo": "bar"}}}}
rss := []*extensions.ReplicaSet{{Spec: extensions.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}}}}}
rss := []*apps.ReplicaSet{{Spec: apps.ReplicaSetSpec{Selector: &metav1.LabelSelector{MatchLabels: map[string]string{"foo": "bar"}}}}}
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
nodeNameToInfo := schedulercache.CreateNodeNameToInfoMap(test.pods, makeLabeledNodeList(test.nodes))
zoneSpread := ServiceAntiAffinity{podLister: schedulertesting.FakePodLister(test.pods), serviceLister: schedulertesting.FakeServiceLister(test.services), label: "zone"}
mataDataProducer := NewPriorityMetadataFactory(
metaDataProducer := NewPriorityMetadataFactory(
schedulertesting.FakeServiceLister(test.services),
schedulertesting.FakeControllerLister(rcs),
schedulertesting.FakeReplicaSetLister(rss),
schedulertesting.FakeStatefulSetLister(sss))
mataData := mataDataProducer(test.pod, nodeNameToInfo)
ttp := priorityFunction(zoneSpread.CalculateAntiAffinityPriorityMap, zoneSpread.CalculateAntiAffinityPriorityReduce, mataData)
metaData := metaDataProducer(test.pod, nodeNameToInfo)
ttp := priorityFunction(zoneSpread.CalculateAntiAffinityPriorityMap, zoneSpread.CalculateAntiAffinityPriorityReduce, metaData)
list, err := ttp(test.pod, nodeNameToInfo, makeLabeledNodeList(test.nodes))
if err != nil {
t.Errorf("unexpected error: %v", err)

View File

@ -41,18 +41,18 @@ func makeNode(node string, milliCPU, memory int64) *v1.Node {
}
}
func priorityFunction(mapFn algorithm.PriorityMapFunction, reduceFn algorithm.PriorityReduceFunction, mataData interface{}) algorithm.PriorityFunction {
func priorityFunction(mapFn algorithm.PriorityMapFunction, reduceFn algorithm.PriorityReduceFunction, metaData interface{}) algorithm.PriorityFunction {
return func(pod *v1.Pod, nodeNameToInfo map[string]*schedulercache.NodeInfo, nodes []*v1.Node) (schedulerapi.HostPriorityList, error) {
result := make(schedulerapi.HostPriorityList, 0, len(nodes))
for i := range nodes {
hostResult, err := mapFn(pod, mataData, nodeNameToInfo[nodes[i].Name])
hostResult, err := mapFn(pod, metaData, nodeNameToInfo[nodes[i].Name])
if err != nil {
return nil, err
}
result = append(result, hostResult)
}
if reduceFn != nil {
if err := reduceFn(pod, mataData, nodeNameToInfo, result); err != nil {
if err := reduceFn(pod, metaData, nodeNameToInfo, result); err != nil {
return nil, err
}
}

View File

@ -11,17 +11,16 @@ go_test(
srcs = [
"non_zero_test.go",
"topologies_test.go",
"util_test.go",
],
embed = [":go_default_library"],
deps = [
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/selection:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/github.com/stretchr/testify/assert:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/selection:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
],
)
@ -30,14 +29,12 @@ go_library(
srcs = [
"non_zero.go",
"topologies.go",
"util.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/algorithm/priorities/util",
deps = [
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
],
)

View File

@ -26,7 +26,7 @@ import (
)
func TestGetNonzeroRequests(t *testing.T) {
tds := []struct {
tests := []struct {
name string
requests v1.ResourceList
expectedCPU int64
@ -65,9 +65,11 @@ func TestGetNonzeroRequests(t *testing.T) {
},
}
for _, td := range tds {
realCPU, realMemory := GetNonzeroRequests(&td.requests)
assert.EqualValuesf(t, td.expectedCPU, realCPU, "Failed to test: %s", td.name)
assert.EqualValuesf(t, td.expectedMemory, realMemory, "Failed to test: %s", td.name)
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
realCPU, realMemory := GetNonzeroRequests(&test.requests)
assert.EqualValuesf(t, test.expectedCPU, realCPU, "Failed to test: %s", test.name)
assert.EqualValuesf(t, test.expectedMemory, realMemory, "Failed to test: %s", test.name)
})
}
}

View File

@ -59,8 +59,10 @@ func TestGetNamespacesFromPodAffinityTerm(t *testing.T) {
}
for _, test := range tests {
realValue := GetNamespacesFromPodAffinityTerm(fakePod(), test.podAffinityTerm)
assert.EqualValuesf(t, test.expectedValue, realValue, "Failed to test: %s", test.name)
t.Run(test.name, func(t *testing.T) {
realValue := GetNamespacesFromPodAffinityTerm(fakePod(), test.podAffinityTerm)
assert.EqualValuesf(t, test.expectedValue, realValue, "Failed to test: %s", test.name)
})
}
}
@ -96,12 +98,14 @@ func TestPodMatchesTermsNamespaceAndSelector(t *testing.T) {
}
for _, test := range tests {
fakeTestPod := fakePod()
fakeTestPod.Namespace = test.podNamespaces
fakeTestPod.Labels = test.podLabels
t.Run(test.name, func(t *testing.T) {
fakeTestPod := fakePod()
fakeTestPod.Namespace = test.podNamespaces
fakeTestPod.Labels = test.podLabels
realValue := PodMatchesTermsNamespaceAndSelector(fakeTestPod, fakeNamespaces, fakeSelector)
assert.EqualValuesf(t, test.expectedResult, realValue, "Faild to test: %s", test.name)
realValue := PodMatchesTermsNamespaceAndSelector(fakeTestPod, fakeNamespaces, fakeSelector)
assert.EqualValuesf(t, test.expectedResult, realValue, "Faild to test: %s", test.name)
})
}
}
@ -248,7 +252,9 @@ func TestNodesHaveSameTopologyKey(t *testing.T) {
}
for _, test := range tests {
got := NodesHaveSameTopologyKey(test.nodeA, test.nodeB, test.topologyKey)
assert.Equalf(t, test.expected, got, "Failed to test: %s", test.name)
t.Run(test.name, func(t *testing.T) {
got := NodesHaveSameTopologyKey(test.nodeA, test.nodeB, test.topologyKey)
assert.Equalf(t, test.expected, got, "Failed to test: %s", test.name)
})
}
}

View File

@ -1,119 +0,0 @@
/*
Copyright 2017 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 util
import (
"testing"
"github.com/stretchr/testify/assert"
"k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
)
func TestGetControllerRef(t *testing.T) {
fakeBlockOwnerDeletion := true
fakeFalseController := false
fakeTrueController := true
fakeEmptyOwnerReference := metav1.OwnerReference{}
tds := []struct {
name string
pod v1.Pod
expectedNil bool
expectedOR metav1.OwnerReference
}{
{
"ownerreference_not_exist",
v1.Pod{},
true,
fakeEmptyOwnerReference,
},
{
"ownerreference_controller_is_nil",
v1.Pod{
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{
{
APIVersion: "extensions/v1beta1",
Kind: "ReplicaSet",
Name: "or-unit-test-5b9cffccff",
UID: "a46372ea-b254-11e7-8373-fa163e25bfb5",
BlockOwnerDeletion: &fakeBlockOwnerDeletion,
},
},
},
},
true,
fakeEmptyOwnerReference,
},
{
"ownerreference_controller_is_false",
v1.Pod{
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{
{
APIVersion: "extensions/v1beta1",
Kind: "ReplicaSet",
Name: "or-unit-test-5b9cffccff",
UID: "a46372ea-b254-11e7-8373-fa163e25bfb5",
Controller: &fakeFalseController,
BlockOwnerDeletion: &fakeBlockOwnerDeletion,
},
},
},
},
true,
fakeEmptyOwnerReference,
},
{
"ownerreference_controller_is_true",
v1.Pod{
ObjectMeta: metav1.ObjectMeta{
OwnerReferences: []metav1.OwnerReference{
{
APIVersion: "extensions/v1beta1",
Kind: "ReplicaSet",
Name: "or-unit-test-5b9cffccff",
UID: "a46372ea-b254-11e7-8373-fa163e25bfb5",
BlockOwnerDeletion: &fakeBlockOwnerDeletion,
Controller: &fakeTrueController,
},
},
},
},
false,
metav1.OwnerReference{
APIVersion: "extensions/v1beta1",
Kind: "ReplicaSet",
Name: "or-unit-test-5b9cffccff",
UID: "a46372ea-b254-11e7-8373-fa163e25bfb5",
BlockOwnerDeletion: &fakeBlockOwnerDeletion,
Controller: &fakeTrueController,
},
},
}
for _, td := range tds {
realOR := GetControllerRef(&td.pod)
if td.expectedNil {
assert.Nilf(t, realOR, "Failed to test: %s", td.name)
} else {
assert.Equalf(t, &td.expectedOR, realOR, "Failed to test: %s", td.name)
}
}
}

View File

@ -26,6 +26,9 @@ import (
// decisions made by Kubernetes. This is typically needed for resources not directly
// managed by Kubernetes.
type SchedulerExtender interface {
// Name returns a unique name that identifies the extender.
Name() string
// Filter based on extender-implemented predicate functions. The filtered list is
// expected to be a subset of the supplied list. failedNodesMap optionally contains
// the list of failed nodes and failure reasons.

View File

@ -1,60 +0,0 @@
/*
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 algorithm
import (
"testing"
"k8s.io/api/core/v1"
)
// Some functions used by multiple scheduler tests.
type schedulerTester struct {
t *testing.T
scheduler ScheduleAlgorithm
nodeLister NodeLister
}
// Call if you know exactly where pod should get scheduled.
func (st *schedulerTester) expectSchedule(pod *v1.Pod, expected string) {
actual, err := st.scheduler.Schedule(pod, st.nodeLister)
if err != nil {
st.t.Errorf("Unexpected error %v\nTried to schedule: %#v", err, pod)
return
}
if actual != expected {
st.t.Errorf("Unexpected scheduling value: %v, expected %v", actual, expected)
}
}
// Call if you can't predict where pod will be scheduled.
func (st *schedulerTester) expectSuccess(pod *v1.Pod) {
_, err := st.scheduler.Schedule(pod, st.nodeLister)
if err != nil {
st.t.Errorf("Unexpected error %v\nTried to schedule: %#v", err, pod)
return
}
}
// Call if pod should *not* schedule.
func (st *schedulerTester) expectFailure(pod *v1.Pod) {
_, err := st.scheduler.Schedule(pod, st.nodeLister)
if err == nil {
st.t.Error("Unexpected non-error")
}
}

View File

@ -17,18 +17,19 @@ limitations under the License.
package algorithm
import (
apps "k8s.io/api/apps/v1beta1"
apps "k8s.io/api/apps/v1"
"k8s.io/api/core/v1"
extensions "k8s.io/api/extensions/v1beta1"
policyv1beta1 "k8s.io/api/policy/v1beta1"
"k8s.io/apimachinery/pkg/labels"
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
schedulerinternalcache "k8s.io/kubernetes/pkg/scheduler/internal/cache"
)
// NodeFieldSelectorKeys is a map that: the key are node field selector keys; the values are
// the functions to get the value of the node field.
var NodeFieldSelectorKeys = map[string]func(*v1.Node) string{
NodeFieldSelectorKeyNodeName: func(n *v1.Node) string { return n.Name },
schedulerapi.NodeFieldSelectorKeyNodeName: func(n *v1.Node) string { return n.Name },
}
// FitPredicate is a function that indicates if a pod fits into an existing node.
@ -98,7 +99,7 @@ type PodLister interface {
List(labels.Selector) ([]*v1.Pod, error)
// This is similar to "List()", but the returned slice does not
// contain pods that don't pass `podFilter`.
FilteredList(podFilter schedulercache.PodFilter, selector labels.Selector) ([]*v1.Pod, error)
FilteredList(podFilter schedulerinternalcache.PodFilter, selector labels.Selector) ([]*v1.Pod, error)
}
// ServiceLister interface represents anything that can produce a list of services; the list is consumed by a scheduler.
@ -120,7 +121,13 @@ type ControllerLister interface {
// ReplicaSetLister interface represents anything that can produce a list of ReplicaSet; the list is consumed by a scheduler.
type ReplicaSetLister interface {
// Gets the replicasets for the given pod
GetPodReplicaSets(*v1.Pod) ([]*extensions.ReplicaSet, error)
GetPodReplicaSets(*v1.Pod) ([]*apps.ReplicaSet, error)
}
// PDBLister interface represents anything that can list PodDisruptionBudget objects.
type PDBLister interface {
// List() returns a list of PodDisruptionBudgets matching the selector.
List(labels.Selector) ([]*policyv1beta1.PodDisruptionBudget, error)
}
var _ ControllerLister = &EmptyControllerLister{}
@ -144,7 +151,7 @@ var _ ReplicaSetLister = &EmptyReplicaSetLister{}
type EmptyReplicaSetLister struct{}
// GetPodReplicaSets returns nil
func (f EmptyReplicaSetLister) GetPodReplicaSets(pod *v1.Pod) (rss []*extensions.ReplicaSet, err error) {
func (f EmptyReplicaSetLister) GetPodReplicaSets(pod *v1.Pod) (rss []*apps.ReplicaSet, err error) {
return nil, nil
}

View File

@ -15,11 +15,16 @@ go_library(
go_test(
name = "go_default_test",
srcs = ["plugins_test.go"],
srcs = [
"main_test.go",
"plugins_test.go",
],
embed = [":go_default_library"],
deps = [
"//pkg/features:go_default_library",
"//pkg/scheduler/factory:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature/testing:go_default_library",
],
)

View File

@ -17,9 +17,9 @@ go_library(
"//pkg/scheduler/algorithm/priorities:go_default_library",
"//pkg/scheduler/core:go_default_library",
"//pkg/scheduler/factory:go_default_library",
"//vendor/github.com/golang/glog:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
@ -36,14 +36,14 @@ go_test(
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/api/latest:go_default_library",
"//pkg/scheduler/factory:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/client-go/informers:go_default_library",
"//vendor/k8s.io/client-go/kubernetes:go_default_library",
"//vendor/k8s.io/client-go/rest:go_default_library",
"//vendor/k8s.io/client-go/util/testing:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/client-go/informers:go_default_library",
"//staging/src/k8s.io/client-go/kubernetes:go_default_library",
"//staging/src/k8s.io/client-go/rest:go_default_library",
"//staging/src/k8s.io/client-go/util/testing:go_default_library",
],
)

View File

@ -814,6 +814,130 @@ func TestCompatibility_v1_Scheduler(t *testing.T) {
}},
},
},
// Do not change this JSON after the corresponding release has been tagged.
// A failure indicates backwards compatibility with the specified release was broken.
"1.12": {
JSON: `{
"kind": "Policy",
"apiVersion": "v1",
"predicates": [
{"name": "MatchNodeSelector"},
{"name": "PodFitsResources"},
{"name": "PodFitsHostPorts"},
{"name": "HostName"},
{"name": "NoDiskConflict"},
{"name": "NoVolumeZoneConflict"},
{"name": "PodToleratesNodeTaints"},
{"name": "CheckNodeMemoryPressure"},
{"name": "CheckNodeDiskPressure"},
{"name": "CheckNodePIDPressure"},
{"name": "CheckNodeCondition"},
{"name": "MaxEBSVolumeCount"},
{"name": "MaxGCEPDVolumeCount"},
{"name": "MaxAzureDiskVolumeCount"},
{"name": "MaxCSIVolumeCountPred"},
{"name": "MatchInterPodAffinity"},
{"name": "GeneralPredicates"},
{"name": "CheckVolumeBinding"},
{"name": "TestServiceAffinity", "argument": {"serviceAffinity" : {"labels" : ["region"]}}},
{"name": "TestLabelsPresence", "argument": {"labelsPresence" : {"labels" : ["foo"], "presence":true}}}
],"priorities": [
{"name": "EqualPriority", "weight": 2},
{"name": "ImageLocalityPriority", "weight": 2},
{"name": "LeastRequestedPriority", "weight": 2},
{"name": "BalancedResourceAllocation", "weight": 2},
{"name": "SelectorSpreadPriority", "weight": 2},
{"name": "NodePreferAvoidPodsPriority", "weight": 2},
{"name": "NodeAffinityPriority", "weight": 2},
{"name": "TaintTolerationPriority", "weight": 2},
{"name": "InterPodAffinityPriority", "weight": 2},
{"name": "MostRequestedPriority", "weight": 2},
{
"name": "RequestedToCapacityRatioPriority",
"weight": 2,
"argument": {
"requestedToCapacityRatioArguments": {
"shape": [
{"utilization": 0, "score": 0},
{"utilization": 50, "score": 7}
]
}
}}
],"extenders": [{
"urlPrefix": "/prefix",
"filterVerb": "filter",
"prioritizeVerb": "prioritize",
"weight": 1,
"bindVerb": "bind",
"enableHttps": true,
"tlsConfig": {"Insecure":true},
"httpTimeout": 1,
"nodeCacheCapable": true,
"managedResources": [{"name":"example.com/foo","ignoredByScheduler":true}],
"ignorable":true
}]
}`,
ExpectedPolicy: schedulerapi.Policy{
Predicates: []schedulerapi.PredicatePolicy{
{Name: "MatchNodeSelector"},
{Name: "PodFitsResources"},
{Name: "PodFitsHostPorts"},
{Name: "HostName"},
{Name: "NoDiskConflict"},
{Name: "NoVolumeZoneConflict"},
{Name: "PodToleratesNodeTaints"},
{Name: "CheckNodeMemoryPressure"},
{Name: "CheckNodeDiskPressure"},
{Name: "CheckNodePIDPressure"},
{Name: "CheckNodeCondition"},
{Name: "MaxEBSVolumeCount"},
{Name: "MaxGCEPDVolumeCount"},
{Name: "MaxAzureDiskVolumeCount"},
{Name: "MaxCSIVolumeCountPred"},
{Name: "MatchInterPodAffinity"},
{Name: "GeneralPredicates"},
{Name: "CheckVolumeBinding"},
{Name: "TestServiceAffinity", Argument: &schedulerapi.PredicateArgument{ServiceAffinity: &schedulerapi.ServiceAffinity{Labels: []string{"region"}}}},
{Name: "TestLabelsPresence", Argument: &schedulerapi.PredicateArgument{LabelsPresence: &schedulerapi.LabelsPresence{Labels: []string{"foo"}, Presence: true}}},
},
Priorities: []schedulerapi.PriorityPolicy{
{Name: "EqualPriority", Weight: 2},
{Name: "ImageLocalityPriority", Weight: 2},
{Name: "LeastRequestedPriority", Weight: 2},
{Name: "BalancedResourceAllocation", Weight: 2},
{Name: "SelectorSpreadPriority", Weight: 2},
{Name: "NodePreferAvoidPodsPriority", Weight: 2},
{Name: "NodeAffinityPriority", Weight: 2},
{Name: "TaintTolerationPriority", Weight: 2},
{Name: "InterPodAffinityPriority", Weight: 2},
{Name: "MostRequestedPriority", Weight: 2},
{
Name: "RequestedToCapacityRatioPriority",
Weight: 2,
Argument: &schedulerapi.PriorityArgument{
RequestedToCapacityRatioArguments: &schedulerapi.RequestedToCapacityRatioArguments{
UtilizationShape: []schedulerapi.UtilizationShapePoint{
{Utilization: 0, Score: 0},
{Utilization: 50, Score: 7},
}},
},
},
},
ExtenderConfigs: []schedulerapi.ExtenderConfig{{
URLPrefix: "/prefix",
FilterVerb: "filter",
PrioritizeVerb: "prioritize",
Weight: 1,
BindVerb: "bind", // 1.11 restored case-sensitivity, but allowed either "BindVerb" or "bindVerb"
EnableHTTPS: true,
TLSConfig: &restclient.TLSClientConfig{Insecure: true},
HTTPTimeout: 1,
NodeCacheCapable: true,
ManagedResources: []schedulerapi.ExtenderManagedResource{{Name: v1.ResourceName("example.com/foo"), IgnoredByScheduler: true}},
Ignorable: true,
}},
},
},
}
registeredPredicates := sets.NewString(factory.ListRegisteredFitPredicates()...)
@ -849,23 +973,24 @@ func TestCompatibility_v1_Scheduler(t *testing.T) {
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
informerFactory := informers.NewSharedInformerFactory(client, 0)
if _, err := factory.NewConfigFactory(
"some-scheduler-name",
client,
informerFactory.Core().V1().Nodes(),
informerFactory.Core().V1().Pods(),
informerFactory.Core().V1().PersistentVolumes(),
informerFactory.Core().V1().PersistentVolumeClaims(),
informerFactory.Core().V1().ReplicationControllers(),
informerFactory.Extensions().V1beta1().ReplicaSets(),
informerFactory.Apps().V1beta1().StatefulSets(),
informerFactory.Core().V1().Services(),
informerFactory.Policy().V1beta1().PodDisruptionBudgets(),
informerFactory.Storage().V1().StorageClasses(),
v1.DefaultHardPodAffinitySymmetricWeight,
enableEquivalenceCache,
false,
).CreateFromConfig(policy); err != nil {
if _, err := factory.NewConfigFactory(&factory.ConfigFactoryArgs{
SchedulerName: "some-scheduler-name",
Client: client,
NodeInformer: informerFactory.Core().V1().Nodes(),
PodInformer: informerFactory.Core().V1().Pods(),
PvInformer: informerFactory.Core().V1().PersistentVolumes(),
PvcInformer: informerFactory.Core().V1().PersistentVolumeClaims(),
ReplicationControllerInformer: informerFactory.Core().V1().ReplicationControllers(),
ReplicaSetInformer: informerFactory.Apps().V1().ReplicaSets(),
StatefulSetInformer: informerFactory.Apps().V1().StatefulSets(),
ServiceInformer: informerFactory.Core().V1().Services(),
PdbInformer: informerFactory.Policy().V1beta1().PodDisruptionBudgets(),
StorageClassInformer: informerFactory.Storage().V1().StorageClasses(),
HardPodAffinitySymmetricWeight: v1.DefaultHardPodAffinitySymmetricWeight,
EnableEquivalenceClassCache: enableEquivalenceCache,
DisablePreemption: false,
PercentageOfNodesToScore: schedulerapi.DefaultPercentageOfNodesToScore,
}).CreateFromConfig(policy); err != nil {
t.Errorf("%s: Error constructing: %v", v, err)
continue
}

View File

@ -17,7 +17,7 @@ limitations under the License.
package defaults
import (
"github.com/golang/glog"
"k8s.io/klog"
"k8s.io/apimachinery/pkg/util/sets"
utilfeature "k8s.io/apiserver/pkg/util/feature"
@ -94,10 +94,6 @@ func init() {
// Register the priority function so that its available
// but do not include it as part of the default priorities
factory.RegisterPriorityFunction2("EqualPriority", core.EqualPriorityMap, nil, 1)
// ImageLocalityPriority prioritizes nodes based on locality of images requested by a pod. Nodes with larger size
// of already-installed packages required by the pod will be preferred over nodes with no already-installed
// packages required by the pod or a small total size of already-installed packages required by the pod.
factory.RegisterPriorityFunction2("ImageLocalityPriority", priorities.ImageLocalityPriorityMap, nil, 1)
// Optional, cluster-autoscaler friendly priority function - give used nodes higher priority.
factory.RegisterPriorityFunction2("MostRequestedPriority", priorities.MostRequestedPriorityMap, nil, 1)
factory.RegisterPriorityFunction2(
@ -137,6 +133,12 @@ func defaultPredicates() sets.String {
return predicates.NewMaxPDVolumeCountPredicate(predicates.AzureDiskVolumeFilterType, args.PVInfo, args.PVCInfo)
},
),
factory.RegisterFitPredicateFactory(
predicates.MaxCSIVolumeCountPred,
func(args factory.PluginFactoryArgs) algorithm.FitPredicate {
return predicates.NewCSIMaxVolumeLimitPredicate(args.PVInfo, args.PVCInfo)
},
),
// Fit is determined by inter-pod affinity.
factory.RegisterFitPredicateFactory(
predicates.MatchInterPodAffinityPred,
@ -197,19 +199,24 @@ func ApplyFeatureGates() {
// Fit is determined based on whether a pod can tolerate all of the node's taints
factory.RegisterMandatoryFitPredicate(predicates.PodToleratesNodeTaintsPred, predicates.PodToleratesNodeTaints)
// Fit is determined based on whether a pod can tolerate unschedulable of node
factory.RegisterMandatoryFitPredicate(predicates.CheckNodeUnschedulablePred, predicates.CheckNodeUnschedulablePredicate)
// Insert Key "PodToleratesNodeTaints" and "CheckNodeUnschedulable" To All Algorithm Provider
// The key will insert to all providers which in algorithmProviderMap[]
// if you just want insert to specific provider, call func InsertPredicateKeyToAlgoProvider()
factory.InsertPredicateKeyToAlgorithmProviderMap(predicates.PodToleratesNodeTaintsPred)
factory.InsertPredicateKeyToAlgorithmProviderMap(predicates.CheckNodeUnschedulablePred)
glog.Warningf("TaintNodesByCondition is enabled, PodToleratesNodeTaints predicate is mandatory")
klog.Infof("TaintNodesByCondition is enabled, PodToleratesNodeTaints predicate is mandatory")
}
// Prioritizes nodes that satisfy pod's resource limits
if utilfeature.DefaultFeatureGate.Enabled(features.ResourceLimitsPriorityFunction) {
klog.Infof("Registering resourcelimits priority function")
factory.RegisterPriorityFunction2("ResourceLimitsPriority", priorities.ResourceLimitsPriorityMap, nil, 1)
// Register the priority function to specific provider too.
factory.InsertPriorityKeyToAlgorithmProviderMap(factory.RegisterPriorityFunction2("ResourceLimitsPriority", priorities.ResourceLimitsPriorityMap, nil, 1))
}
}
func registerAlgorithmProvider(predSet, priSet sets.String) {
@ -260,6 +267,9 @@ func defaultPriorities() sets.String {
// Prioritizes nodes that marked with taint which pod can tolerate.
factory.RegisterPriorityFunction2("TaintTolerationPriority", priorities.ComputeTaintTolerationPriorityMap, priorities.ComputeTaintTolerationPriorityReduce, 1),
// ImageLocalityPriority prioritizes nodes that have images requested by the pod present.
factory.RegisterPriorityFunction2("ImageLocalityPriority", priorities.ImageLocalityPriorityMap, nil, 1),
)
}

View File

@ -59,7 +59,8 @@ func TestDefaultPriorities(t *testing.T) {
"BalancedResourceAllocation",
"NodePreferAvoidPodsPriority",
"NodeAffinityPriority",
"TaintTolerationPriority")
"TaintTolerationPriority",
"ImageLocalityPriority")
if expected := defaultPriorities(); !result.Equal(expected) {
t.Errorf("expected %v got %v", expected, result)
}
@ -67,18 +68,19 @@ func TestDefaultPriorities(t *testing.T) {
func TestDefaultPredicates(t *testing.T) {
result := sets.NewString(
"NoVolumeZoneConflict",
"MaxEBSVolumeCount",
"MaxGCEPDVolumeCount",
"MaxAzureDiskVolumeCount",
"MatchInterPodAffinity",
"NoDiskConflict",
"GeneralPredicates",
"CheckNodeMemoryPressure",
"CheckNodeDiskPressure",
"CheckNodePIDPressure",
"CheckNodeCondition",
"PodToleratesNodeTaints",
predicates.NoVolumeZoneConflictPred,
predicates.MaxEBSVolumeCountPred,
predicates.MaxGCEPDVolumeCountPred,
predicates.MaxAzureDiskVolumeCountPred,
predicates.MaxCSIVolumeCountPred,
predicates.MatchInterPodAffinityPred,
predicates.NoDiskConflictPred,
predicates.GeneralPred,
predicates.CheckNodeMemoryPressurePred,
predicates.CheckNodeDiskPressurePred,
predicates.CheckNodePIDPressurePred,
predicates.CheckNodeConditionPred,
predicates.PodToleratesNodeTaintsPred,
predicates.CheckVolumeBindingPred,
)

View File

@ -0,0 +1,29 @@
/*
Copyright 2018 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 algorithmprovider
import (
"testing"
utilfeature "k8s.io/apiserver/pkg/util/feature"
utilfeaturetesting "k8s.io/apiserver/pkg/util/feature/testing"
_ "k8s.io/kubernetes/pkg/features"
)
func TestMain(m *testing.M) {
utilfeaturetesting.VerifyFeatureGatesUnchanged(utilfeature.DefaultFeatureGate, m.Run)
}

View File

@ -17,9 +17,12 @@ limitations under the License.
package algorithmprovider
import (
"fmt"
"testing"
utilfeature "k8s.io/apiserver/pkg/util/feature"
utilfeaturetesting "k8s.io/apiserver/pkg/util/feature/testing"
"k8s.io/kubernetes/pkg/features"
"k8s.io/kubernetes/pkg/scheduler/factory"
)
@ -44,66 +47,69 @@ func TestDefaultConfigExists(t *testing.T) {
func TestAlgorithmProviders(t *testing.T) {
for _, pn := range algorithmProviderNames {
p, err := factory.GetAlgorithmProvider(pn)
if err != nil {
t.Errorf("error retrieving '%s' provider: %v", pn, err)
break
}
if len(p.PriorityFunctionKeys) == 0 {
t.Errorf("%s algorithm provider shouldn't have 0 priority functions", pn)
}
for _, pf := range p.PriorityFunctionKeys.List() {
if !factory.IsPriorityFunctionRegistered(pf) {
t.Errorf("priority function %s is not registered but is used in the %s algorithm provider", pf, pn)
t.Run(pn, func(t *testing.T) {
p, err := factory.GetAlgorithmProvider(pn)
if err != nil {
t.Fatalf("error retrieving provider: %v", err)
}
}
for _, fp := range p.FitPredicateKeys.List() {
if !factory.IsFitPredicateRegistered(fp) {
t.Errorf("fit predicate %s is not registered but is used in the %s algorithm provider", fp, pn)
if len(p.PriorityFunctionKeys) == 0 {
t.Errorf("algorithm provider shouldn't have 0 priority functions")
}
}
for _, pf := range p.PriorityFunctionKeys.List() {
t.Run(fmt.Sprintf("priorityfunction/%s", pf), func(t *testing.T) {
if !factory.IsPriorityFunctionRegistered(pf) {
t.Errorf("priority function is not registered but is used in the algorithm provider")
}
})
}
for _, fp := range p.FitPredicateKeys.List() {
t.Run(fmt.Sprintf("fitpredicate/%s", fp), func(t *testing.T) {
if !factory.IsFitPredicateRegistered(fp) {
t.Errorf("fit predicate is not registered but is used in the algorithm provider")
}
})
}
})
}
}
func TestApplyFeatureGates(t *testing.T) {
for _, pn := range algorithmProviderNames {
p, err := factory.GetAlgorithmProvider(pn)
if err != nil {
t.Errorf("Error retrieving '%s' provider: %v", pn, err)
break
}
t.Run(pn, func(t *testing.T) {
p, err := factory.GetAlgorithmProvider(pn)
if err != nil {
t.Fatalf("Error retrieving provider: %v", err)
}
if !p.FitPredicateKeys.Has("CheckNodeCondition") {
t.Errorf("Failed to find predicate: 'CheckNodeCondition'")
break
}
if !p.FitPredicateKeys.Has("CheckNodeCondition") {
t.Fatalf("Failed to find predicate: 'CheckNodeCondition'")
}
if !p.FitPredicateKeys.Has("PodToleratesNodeTaints") {
t.Errorf("Failed to find predicate: 'PodToleratesNodeTaints'")
break
}
if !p.FitPredicateKeys.Has("PodToleratesNodeTaints") {
t.Fatalf("Failed to find predicate: 'PodToleratesNodeTaints'")
}
})
}
// Apply features for algorithm providers.
utilfeature.DefaultFeatureGate.Set("TaintNodesByCondition=True")
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.TaintNodesByCondition, true)()
ApplyFeatureGates()
for _, pn := range algorithmProviderNames {
p, err := factory.GetAlgorithmProvider(pn)
if err != nil {
t.Errorf("Error retrieving '%s' provider: %v", pn, err)
break
}
t.Run(pn, func(t *testing.T) {
p, err := factory.GetAlgorithmProvider(pn)
if err != nil {
t.Fatalf("Error retrieving '%s' provider: %v", pn, err)
}
if !p.FitPredicateKeys.Has("PodToleratesNodeTaints") {
t.Errorf("Failed to find predicate: 'PodToleratesNodeTaints'")
break
}
if !p.FitPredicateKeys.Has("PodToleratesNodeTaints") {
t.Fatalf("Failed to find predicate: 'PodToleratesNodeTaints'")
}
if p.FitPredicateKeys.Has("CheckNodeCondition") {
t.Errorf("Unexpected predicate: 'CheckNodeCondition'")
break
}
if p.FitPredicateKeys.Has("CheckNodeCondition") {
t.Fatalf("Unexpected predicate: 'CheckNodeCondition'")
}
})
}
}

View File

@ -11,16 +11,18 @@ go_library(
"doc.go",
"register.go",
"types.go",
"well_known_labels.go",
"zz_generated.deepcopy.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/api",
deps = [
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/types:go_default_library",
"//vendor/k8s.io/client-go/rest:go_default_library",
"//pkg/apis/core:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/types:go_default_library",
"//staging/src/k8s.io/client-go/rest:go_default_library",
],
)

View File

@ -12,10 +12,10 @@ go_library(
deps = [
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/api/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime/serializer/json:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime/serializer/versioning:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/serializer/json:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/serializer/versioning:go_default_library",
],
)

View File

@ -36,6 +36,9 @@ const (
MaxPriority = 10
// MaxWeight defines the max weight value.
MaxWeight = MaxInt / MaxPriority
// DefaultPercentageOfNodesToScore defines the percentage of nodes of all nodes
// that once found feasible, the scheduler stops looking for more nodes.
DefaultPercentageOfNodesToScore = 50
)
// +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object

View File

@ -16,12 +16,12 @@ go_library(
importpath = "k8s.io/kubernetes/pkg/scheduler/api/v1",
deps = [
"//pkg/scheduler/api:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/types:go_default_library",
"//vendor/k8s.io/client-go/rest:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/types:go_default_library",
"//staging/src/k8s.io/client-go/rest:go_default_library",
],
)

View File

@ -21,7 +21,7 @@ limitations under the License.
package v1
import (
core_v1 "k8s.io/api/core/v1"
corev1 "k8s.io/api/core/v1"
runtime "k8s.io/apimachinery/pkg/runtime"
rest "k8s.io/client-go/rest"
)
@ -31,33 +31,21 @@ func (in *ExtenderArgs) DeepCopyInto(out *ExtenderArgs) {
*out = *in
if in.Pod != nil {
in, out := &in.Pod, &out.Pod
if *in == nil {
*out = nil
} else {
*out = new(core_v1.Pod)
(*in).DeepCopyInto(*out)
}
*out = new(corev1.Pod)
(*in).DeepCopyInto(*out)
}
if in.Nodes != nil {
in, out := &in.Nodes, &out.Nodes
if *in == nil {
*out = nil
} else {
*out = new(core_v1.NodeList)
(*in).DeepCopyInto(*out)
}
*out = new(corev1.NodeList)
(*in).DeepCopyInto(*out)
}
if in.NodeNames != nil {
in, out := &in.NodeNames, &out.NodeNames
if *in == nil {
*out = nil
} else {
*out = new([]string)
if **in != nil {
in, out := *in, *out
*out = make([]string, len(*in))
copy(*out, *in)
}
*out = new([]string)
if **in != nil {
in, out := *in, *out
*out = make([]string, len(*in))
copy(*out, *in)
}
}
return
@ -110,12 +98,8 @@ func (in *ExtenderConfig) DeepCopyInto(out *ExtenderConfig) {
*out = *in
if in.TLSConfig != nil {
in, out := &in.TLSConfig, &out.TLSConfig
if *in == nil {
*out = nil
} else {
*out = new(rest.TLSClientConfig)
(*in).DeepCopyInto(*out)
}
*out = new(rest.TLSClientConfig)
(*in).DeepCopyInto(*out)
}
if in.ManagedResources != nil {
in, out := &in.ManagedResources, &out.ManagedResources
@ -140,24 +124,16 @@ func (in *ExtenderFilterResult) DeepCopyInto(out *ExtenderFilterResult) {
*out = *in
if in.Nodes != nil {
in, out := &in.Nodes, &out.Nodes
if *in == nil {
*out = nil
} else {
*out = new(core_v1.NodeList)
(*in).DeepCopyInto(*out)
}
*out = new(corev1.NodeList)
(*in).DeepCopyInto(*out)
}
if in.NodeNames != nil {
in, out := &in.NodeNames, &out.NodeNames
if *in == nil {
*out = nil
} else {
*out = new([]string)
if **in != nil {
in, out := *in, *out
*out = make([]string, len(*in))
copy(*out, *in)
}
*out = new([]string)
if **in != nil {
in, out := *in, *out
*out = make([]string, len(*in))
copy(*out, *in)
}
}
if in.FailedNodes != nil {
@ -201,35 +177,37 @@ func (in *ExtenderPreemptionArgs) DeepCopyInto(out *ExtenderPreemptionArgs) {
*out = *in
if in.Pod != nil {
in, out := &in.Pod, &out.Pod
if *in == nil {
*out = nil
} else {
*out = new(core_v1.Pod)
(*in).DeepCopyInto(*out)
}
*out = new(corev1.Pod)
(*in).DeepCopyInto(*out)
}
if in.NodeNameToVictims != nil {
in, out := &in.NodeNameToVictims, &out.NodeNameToVictims
*out = make(map[string]*Victims, len(*in))
for key, val := range *in {
var outVal *Victims
if val == nil {
(*out)[key] = nil
} else {
(*out)[key] = new(Victims)
val.DeepCopyInto((*out)[key])
in, out := &val, &outVal
*out = new(Victims)
(*in).DeepCopyInto(*out)
}
(*out)[key] = outVal
}
}
if in.NodeNameToMetaVictims != nil {
in, out := &in.NodeNameToMetaVictims, &out.NodeNameToMetaVictims
*out = make(map[string]*MetaVictims, len(*in))
for key, val := range *in {
var outVal *MetaVictims
if val == nil {
(*out)[key] = nil
} else {
(*out)[key] = new(MetaVictims)
val.DeepCopyInto((*out)[key])
in, out := &val, &outVal
*out = new(MetaVictims)
(*in).DeepCopyInto(*out)
}
(*out)[key] = outVal
}
}
return
@ -252,12 +230,15 @@ func (in *ExtenderPreemptionResult) DeepCopyInto(out *ExtenderPreemptionResult)
in, out := &in.NodeNameToMetaVictims, &out.NodeNameToMetaVictims
*out = make(map[string]*MetaVictims, len(*in))
for key, val := range *in {
var outVal *MetaVictims
if val == nil {
(*out)[key] = nil
} else {
(*out)[key] = new(MetaVictims)
val.DeepCopyInto((*out)[key])
in, out := &val, &outVal
*out = new(MetaVictims)
(*in).DeepCopyInto(*out)
}
(*out)[key] = outVal
}
}
return
@ -391,11 +372,10 @@ func (in *MetaVictims) DeepCopyInto(out *MetaVictims) {
in, out := &in.Pods, &out.Pods
*out = make([]*MetaPod, len(*in))
for i := range *in {
if (*in)[i] == nil {
(*out)[i] = nil
} else {
(*out)[i] = new(MetaPod)
(*in)[i].DeepCopyInto((*out)[i])
if (*in)[i] != nil {
in, out := &(*in)[i], &(*out)[i]
*out = new(MetaPod)
**out = **in
}
}
}
@ -463,21 +443,13 @@ func (in *PredicateArgument) DeepCopyInto(out *PredicateArgument) {
*out = *in
if in.ServiceAffinity != nil {
in, out := &in.ServiceAffinity, &out.ServiceAffinity
if *in == nil {
*out = nil
} else {
*out = new(ServiceAffinity)
(*in).DeepCopyInto(*out)
}
*out = new(ServiceAffinity)
(*in).DeepCopyInto(*out)
}
if in.LabelsPresence != nil {
in, out := &in.LabelsPresence, &out.LabelsPresence
if *in == nil {
*out = nil
} else {
*out = new(LabelsPresence)
(*in).DeepCopyInto(*out)
}
*out = new(LabelsPresence)
(*in).DeepCopyInto(*out)
}
return
}
@ -497,12 +469,8 @@ func (in *PredicatePolicy) DeepCopyInto(out *PredicatePolicy) {
*out = *in
if in.Argument != nil {
in, out := &in.Argument, &out.Argument
if *in == nil {
*out = nil
} else {
*out = new(PredicateArgument)
(*in).DeepCopyInto(*out)
}
*out = new(PredicateArgument)
(*in).DeepCopyInto(*out)
}
return
}
@ -522,30 +490,18 @@ func (in *PriorityArgument) DeepCopyInto(out *PriorityArgument) {
*out = *in
if in.ServiceAntiAffinity != nil {
in, out := &in.ServiceAntiAffinity, &out.ServiceAntiAffinity
if *in == nil {
*out = nil
} else {
*out = new(ServiceAntiAffinity)
**out = **in
}
*out = new(ServiceAntiAffinity)
**out = **in
}
if in.LabelPreference != nil {
in, out := &in.LabelPreference, &out.LabelPreference
if *in == nil {
*out = nil
} else {
*out = new(LabelPreference)
**out = **in
}
*out = new(LabelPreference)
**out = **in
}
if in.RequestedToCapacityRatioArguments != nil {
in, out := &in.RequestedToCapacityRatioArguments, &out.RequestedToCapacityRatioArguments
if *in == nil {
*out = nil
} else {
*out = new(RequestedToCapacityRatioArguments)
(*in).DeepCopyInto(*out)
}
*out = new(RequestedToCapacityRatioArguments)
(*in).DeepCopyInto(*out)
}
return
}
@ -565,12 +521,8 @@ func (in *PriorityPolicy) DeepCopyInto(out *PriorityPolicy) {
*out = *in
if in.Argument != nil {
in, out := &in.Argument, &out.Argument
if *in == nil {
*out = nil
} else {
*out = new(PriorityArgument)
(*in).DeepCopyInto(*out)
}
*out = new(PriorityArgument)
(*in).DeepCopyInto(*out)
}
return
}
@ -664,13 +616,12 @@ func (in *Victims) DeepCopyInto(out *Victims) {
*out = *in
if in.Pods != nil {
in, out := &in.Pods, &out.Pods
*out = make([]*core_v1.Pod, len(*in))
*out = make([]*corev1.Pod, len(*in))
for i := range *in {
if (*in)[i] == nil {
(*out)[i] = nil
} else {
(*out)[i] = new(core_v1.Pod)
(*in)[i].DeepCopyInto((*out)[i])
if (*in)[i] != nil {
in, out := &(*in)[i], &(*out)[i]
*out = new(corev1.Pod)
(*in).DeepCopyInto(*out)
}
}
}

View File

@ -13,10 +13,10 @@ go_library(
deps = [
"//pkg/apis/core/v1/helper:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/errors:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/validation:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/errors:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/validation:go_default_library",
],
)

View File

@ -14,7 +14,7 @@ See the License for the specific language governing permissions and
limitations under the License.
*/
package algorithm
package api
import (
api "k8s.io/kubernetes/pkg/apis/core"

View File

@ -31,33 +31,21 @@ func (in *ExtenderArgs) DeepCopyInto(out *ExtenderArgs) {
*out = *in
if in.Pod != nil {
in, out := &in.Pod, &out.Pod
if *in == nil {
*out = nil
} else {
*out = new(v1.Pod)
(*in).DeepCopyInto(*out)
}
*out = new(v1.Pod)
(*in).DeepCopyInto(*out)
}
if in.Nodes != nil {
in, out := &in.Nodes, &out.Nodes
if *in == nil {
*out = nil
} else {
*out = new(v1.NodeList)
(*in).DeepCopyInto(*out)
}
*out = new(v1.NodeList)
(*in).DeepCopyInto(*out)
}
if in.NodeNames != nil {
in, out := &in.NodeNames, &out.NodeNames
if *in == nil {
*out = nil
} else {
*out = new([]string)
if **in != nil {
in, out := *in, *out
*out = make([]string, len(*in))
copy(*out, *in)
}
*out = new([]string)
if **in != nil {
in, out := *in, *out
*out = make([]string, len(*in))
copy(*out, *in)
}
}
return
@ -110,12 +98,8 @@ func (in *ExtenderConfig) DeepCopyInto(out *ExtenderConfig) {
*out = *in
if in.TLSConfig != nil {
in, out := &in.TLSConfig, &out.TLSConfig
if *in == nil {
*out = nil
} else {
*out = new(rest.TLSClientConfig)
(*in).DeepCopyInto(*out)
}
*out = new(rest.TLSClientConfig)
(*in).DeepCopyInto(*out)
}
if in.ManagedResources != nil {
in, out := &in.ManagedResources, &out.ManagedResources
@ -140,24 +124,16 @@ func (in *ExtenderFilterResult) DeepCopyInto(out *ExtenderFilterResult) {
*out = *in
if in.Nodes != nil {
in, out := &in.Nodes, &out.Nodes
if *in == nil {
*out = nil
} else {
*out = new(v1.NodeList)
(*in).DeepCopyInto(*out)
}
*out = new(v1.NodeList)
(*in).DeepCopyInto(*out)
}
if in.NodeNames != nil {
in, out := &in.NodeNames, &out.NodeNames
if *in == nil {
*out = nil
} else {
*out = new([]string)
if **in != nil {
in, out := *in, *out
*out = make([]string, len(*in))
copy(*out, *in)
}
*out = new([]string)
if **in != nil {
in, out := *in, *out
*out = make([]string, len(*in))
copy(*out, *in)
}
}
if in.FailedNodes != nil {
@ -201,35 +177,37 @@ func (in *ExtenderPreemptionArgs) DeepCopyInto(out *ExtenderPreemptionArgs) {
*out = *in
if in.Pod != nil {
in, out := &in.Pod, &out.Pod
if *in == nil {
*out = nil
} else {
*out = new(v1.Pod)
(*in).DeepCopyInto(*out)
}
*out = new(v1.Pod)
(*in).DeepCopyInto(*out)
}
if in.NodeNameToVictims != nil {
in, out := &in.NodeNameToVictims, &out.NodeNameToVictims
*out = make(map[string]*Victims, len(*in))
for key, val := range *in {
var outVal *Victims
if val == nil {
(*out)[key] = nil
} else {
(*out)[key] = new(Victims)
val.DeepCopyInto((*out)[key])
in, out := &val, &outVal
*out = new(Victims)
(*in).DeepCopyInto(*out)
}
(*out)[key] = outVal
}
}
if in.NodeNameToMetaVictims != nil {
in, out := &in.NodeNameToMetaVictims, &out.NodeNameToMetaVictims
*out = make(map[string]*MetaVictims, len(*in))
for key, val := range *in {
var outVal *MetaVictims
if val == nil {
(*out)[key] = nil
} else {
(*out)[key] = new(MetaVictims)
val.DeepCopyInto((*out)[key])
in, out := &val, &outVal
*out = new(MetaVictims)
(*in).DeepCopyInto(*out)
}
(*out)[key] = outVal
}
}
return
@ -252,12 +230,15 @@ func (in *ExtenderPreemptionResult) DeepCopyInto(out *ExtenderPreemptionResult)
in, out := &in.NodeNameToMetaVictims, &out.NodeNameToMetaVictims
*out = make(map[string]*MetaVictims, len(*in))
for key, val := range *in {
var outVal *MetaVictims
if val == nil {
(*out)[key] = nil
} else {
(*out)[key] = new(MetaVictims)
val.DeepCopyInto((*out)[key])
in, out := &val, &outVal
*out = new(MetaVictims)
(*in).DeepCopyInto(*out)
}
(*out)[key] = outVal
}
}
return
@ -391,11 +372,10 @@ func (in *MetaVictims) DeepCopyInto(out *MetaVictims) {
in, out := &in.Pods, &out.Pods
*out = make([]*MetaPod, len(*in))
for i := range *in {
if (*in)[i] == nil {
(*out)[i] = nil
} else {
(*out)[i] = new(MetaPod)
(*in)[i].DeepCopyInto((*out)[i])
if (*in)[i] != nil {
in, out := &(*in)[i], &(*out)[i]
*out = new(MetaPod)
**out = **in
}
}
}
@ -463,21 +443,13 @@ func (in *PredicateArgument) DeepCopyInto(out *PredicateArgument) {
*out = *in
if in.ServiceAffinity != nil {
in, out := &in.ServiceAffinity, &out.ServiceAffinity
if *in == nil {
*out = nil
} else {
*out = new(ServiceAffinity)
(*in).DeepCopyInto(*out)
}
*out = new(ServiceAffinity)
(*in).DeepCopyInto(*out)
}
if in.LabelsPresence != nil {
in, out := &in.LabelsPresence, &out.LabelsPresence
if *in == nil {
*out = nil
} else {
*out = new(LabelsPresence)
(*in).DeepCopyInto(*out)
}
*out = new(LabelsPresence)
(*in).DeepCopyInto(*out)
}
return
}
@ -497,12 +469,8 @@ func (in *PredicatePolicy) DeepCopyInto(out *PredicatePolicy) {
*out = *in
if in.Argument != nil {
in, out := &in.Argument, &out.Argument
if *in == nil {
*out = nil
} else {
*out = new(PredicateArgument)
(*in).DeepCopyInto(*out)
}
*out = new(PredicateArgument)
(*in).DeepCopyInto(*out)
}
return
}
@ -522,30 +490,18 @@ func (in *PriorityArgument) DeepCopyInto(out *PriorityArgument) {
*out = *in
if in.ServiceAntiAffinity != nil {
in, out := &in.ServiceAntiAffinity, &out.ServiceAntiAffinity
if *in == nil {
*out = nil
} else {
*out = new(ServiceAntiAffinity)
**out = **in
}
*out = new(ServiceAntiAffinity)
**out = **in
}
if in.LabelPreference != nil {
in, out := &in.LabelPreference, &out.LabelPreference
if *in == nil {
*out = nil
} else {
*out = new(LabelPreference)
**out = **in
}
*out = new(LabelPreference)
**out = **in
}
if in.RequestedToCapacityRatioArguments != nil {
in, out := &in.RequestedToCapacityRatioArguments, &out.RequestedToCapacityRatioArguments
if *in == nil {
*out = nil
} else {
*out = new(RequestedToCapacityRatioArguments)
(*in).DeepCopyInto(*out)
}
*out = new(RequestedToCapacityRatioArguments)
(*in).DeepCopyInto(*out)
}
return
}
@ -565,12 +521,8 @@ func (in *PriorityPolicy) DeepCopyInto(out *PriorityPolicy) {
*out = *in
if in.Argument != nil {
in, out := &in.Argument, &out.Argument
if *in == nil {
*out = nil
} else {
*out = new(PriorityArgument)
(*in).DeepCopyInto(*out)
}
*out = new(PriorityArgument)
(*in).DeepCopyInto(*out)
}
return
}
@ -666,11 +618,10 @@ func (in *Victims) DeepCopyInto(out *Victims) {
in, out := &in.Pods, &out.Pods
*out = make([]*v1.Pod, len(*in))
for i := range *in {
if (*in)[i] == nil {
(*out)[i] = nil
} else {
(*out)[i] = new(v1.Pod)
(*in)[i].DeepCopyInto((*out)[i])
if (*in)[i] != nil {
in, out := &(*in)[i], &(*out)[i]
*out = new(v1.Pod)
(*in).DeepCopyInto(*out)
}
}
}

View File

@ -0,0 +1,39 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library")
go_library(
name = "go_default_library",
srcs = [
"doc.go",
"register.go",
"types.go",
"zz_generated.deepcopy.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/apis/config",
visibility = ["//visibility:public"],
deps = [
"//staging/src/k8s.io/apimachinery/pkg/apis/config:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/apis/config:go_default_library",
],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [
":package-srcs",
"//pkg/scheduler/apis/config/scheme:all-srcs",
"//pkg/scheduler/apis/config/v1alpha1:all-srcs",
"//pkg/scheduler/apis/config/validation:all-srcs",
],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

11
vendor/k8s.io/kubernetes/pkg/scheduler/apis/config/OWNERS generated vendored Executable file
View File

@ -0,0 +1,11 @@
approvers:
- api-approvers
- sig-scheduling-maintainers
- sttts
- luxas
reviewers:
- sig-scheduling
- api-reviewers
- dixudx
- luxas
- sttts

View File

@ -0,0 +1,20 @@
/*
Copyright 2018 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.
*/
// +k8s:deepcopy-gen=package
// +groupName=kubescheduler.config.k8s.io
package config // import "k8s.io/kubernetes/pkg/scheduler/apis/config"

View File

@ -0,0 +1,43 @@
/*
Copyright 2018 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 config
import (
"k8s.io/apimachinery/pkg/runtime"
"k8s.io/apimachinery/pkg/runtime/schema"
)
// GroupName is the group name used in this package
const GroupName = "kubescheduler.config.k8s.io"
// SchemeGroupVersion is group version used to register these objects
var SchemeGroupVersion = schema.GroupVersion{Group: GroupName, Version: runtime.APIVersionInternal}
var (
// SchemeBuilder is the scheme builder with scheme init functions to run for this API package
SchemeBuilder = runtime.NewSchemeBuilder(addKnownTypes)
// AddToScheme is a global function that registers this API group & version to a scheme
AddToScheme = SchemeBuilder.AddToScheme
)
// addKnownTypes registers known types to the given scheme
func addKnownTypes(scheme *runtime.Scheme) error {
scheme.AddKnownTypes(SchemeGroupVersion,
&KubeSchedulerConfiguration{},
)
return nil
}

View File

@ -0,0 +1,29 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library")
go_library(
name = "go_default_library",
srcs = ["scheme.go"],
importpath = "k8s.io/kubernetes/pkg/scheduler/apis/config/scheme",
visibility = ["//visibility:public"],
deps = [
"//pkg/scheduler/apis/config:go_default_library",
"//pkg/scheduler/apis/config/v1alpha1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/serializer:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/runtime:go_default_library",
],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [":package-srcs"],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

View File

@ -0,0 +1,44 @@
/*
Copyright 2018 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 scheme
import (
"k8s.io/apimachinery/pkg/runtime"
"k8s.io/apimachinery/pkg/runtime/serializer"
utilruntime "k8s.io/apimachinery/pkg/util/runtime"
kubeschedulerconfig "k8s.io/kubernetes/pkg/scheduler/apis/config"
kubeschedulerconfigv1alpha1 "k8s.io/kubernetes/pkg/scheduler/apis/config/v1alpha1"
)
var (
// Scheme is the runtime.Scheme to which all kubescheduler api types are registered.
Scheme = runtime.NewScheme()
// Codecs provides access to encoding and decoding for the scheme.
Codecs = serializer.NewCodecFactory(Scheme)
)
func init() {
AddToScheme(Scheme)
}
// AddToScheme builds the kubescheduler scheme using all known versions of the kubescheduler api.
func AddToScheme(scheme *runtime.Scheme) {
utilruntime.Must(kubeschedulerconfig.AddToScheme(Scheme))
utilruntime.Must(kubeschedulerconfigv1alpha1.AddToScheme(Scheme))
utilruntime.Must(scheme.SetVersionPriority(kubeschedulerconfigv1alpha1.SchemeGroupVersion))
}

View File

@ -0,0 +1,137 @@
/*
Copyright 2018 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 config
import (
apimachineryconfig "k8s.io/apimachinery/pkg/apis/config"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
apiserverconfig "k8s.io/apiserver/pkg/apis/config"
)
const (
// SchedulerDefaultLockObjectNamespace defines default scheduler lock object namespace ("kube-system")
SchedulerDefaultLockObjectNamespace string = metav1.NamespaceSystem
// SchedulerDefaultLockObjectName defines default scheduler lock object name ("kube-scheduler")
SchedulerDefaultLockObjectName = "kube-scheduler"
// SchedulerPolicyConfigMapKey defines the key of the element in the
// scheduler's policy ConfigMap that contains scheduler's policy config.
SchedulerPolicyConfigMapKey = "policy.cfg"
// SchedulerDefaultProviderName defines the default provider names
SchedulerDefaultProviderName = "DefaultProvider"
)
// +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
// KubeSchedulerConfiguration configures a scheduler
type KubeSchedulerConfiguration struct {
metav1.TypeMeta
// SchedulerName is name of the scheduler, used to select which pods
// will be processed by this scheduler, based on pod's "spec.SchedulerName".
SchedulerName string
// AlgorithmSource specifies the scheduler algorithm source.
AlgorithmSource SchedulerAlgorithmSource
// RequiredDuringScheduling affinity is not symmetric, but there is an implicit PreferredDuringScheduling affinity rule
// corresponding to every RequiredDuringScheduling affinity rule.
// HardPodAffinitySymmetricWeight represents the weight of implicit PreferredDuringScheduling affinity rule, in the range 0-100.
HardPodAffinitySymmetricWeight int32
// LeaderElection defines the configuration of leader election client.
LeaderElection KubeSchedulerLeaderElectionConfiguration
// ClientConnection specifies the kubeconfig file and client connection
// settings for the proxy server to use when communicating with the apiserver.
ClientConnection apimachineryconfig.ClientConnectionConfiguration
// HealthzBindAddress is the IP address and port for the health check server to serve on,
// defaulting to 0.0.0.0:10251
HealthzBindAddress string
// MetricsBindAddress is the IP address and port for the metrics server to
// serve on, defaulting to 0.0.0.0:10251.
MetricsBindAddress string
// DebuggingConfiguration holds configuration for Debugging related features
// TODO: We might wanna make this a substruct like Debugging apiserverconfig.DebuggingConfiguration
apiserverconfig.DebuggingConfiguration
// DisablePreemption disables the pod preemption feature.
DisablePreemption bool
// PercentageOfNodeToScore is the percentage of all nodes that once found feasible
// for running a pod, the scheduler stops its search for more feasible nodes in
// the cluster. This helps improve scheduler's performance. Scheduler always tries to find
// at least "minFeasibleNodesToFind" feasible nodes no matter what the value of this flag is.
// Example: if the cluster size is 500 nodes and the value of this flag is 30,
// then scheduler stops finding further feasible nodes once it finds 150 feasible ones.
// When the value is 0, default percentage (50%) of the nodes will be scored.
PercentageOfNodesToScore int32
// DEPRECATED.
// Indicate the "all topologies" set for empty topologyKey when it's used for PreferredDuringScheduling pod anti-affinity.
FailureDomains string
// Duration to wait for a binding operation to complete before timing out
// Value must be non-negative integer. The value zero indicates no waiting.
// If this value is nil, the default value will be used.
BindTimeoutSeconds *int64
}
// SchedulerAlgorithmSource is the source of a scheduler algorithm. One source
// field must be specified, and source fields are mutually exclusive.
type SchedulerAlgorithmSource struct {
// Policy is a policy based algorithm source.
Policy *SchedulerPolicySource
// Provider is the name of a scheduling algorithm provider to use.
Provider *string
}
// SchedulerPolicySource configures a means to obtain a scheduler Policy. One
// source field must be specified, and source fields are mutually exclusive.
type SchedulerPolicySource struct {
// File is a file policy source.
File *SchedulerPolicyFileSource
// ConfigMap is a config map policy source.
ConfigMap *SchedulerPolicyConfigMapSource
}
// SchedulerPolicyFileSource is a policy serialized to disk and accessed via
// path.
type SchedulerPolicyFileSource struct {
// Path is the location of a serialized policy.
Path string
}
// SchedulerPolicyConfigMapSource is a policy serialized into a config map value
// under the SchedulerPolicyConfigMapKey key.
type SchedulerPolicyConfigMapSource struct {
// Namespace is the namespace of the policy config map.
Namespace string
// Name is the name of hte policy config map.
Name string
}
// KubeSchedulerLeaderElectionConfiguration expands LeaderElectionConfiguration
// to include scheduler specific configuration.
type KubeSchedulerLeaderElectionConfiguration struct {
apiserverconfig.LeaderElectionConfiguration
// LockObjectNamespace defines the namespace of the lock object
LockObjectNamespace string
// LockObjectName defines the lock object name
LockObjectName string
}

View File

@ -0,0 +1,53 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
go_library(
name = "go_default_library",
srcs = [
"defaults.go",
"doc.go",
"register.go",
"zz_generated.conversion.go",
"zz_generated.deepcopy.go",
"zz_generated.defaults.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/apis/config/v1alpha1",
visibility = ["//visibility:public"],
deps = [
"//pkg/apis/core:go_default_library",
"//pkg/kubelet/apis:go_default_library",
"//pkg/master/ports:go_default_library",
"//pkg/scheduler/apis/config:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/config/v1alpha1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/conversion:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/apis/config/v1alpha1:go_default_library",
"//staging/src/k8s.io/kube-scheduler/config/v1alpha1:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = ["defaults_test.go"],
embed = [":go_default_library"],
deps = [
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/kube-scheduler/config/v1alpha1:go_default_library",
],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [":package-srcs"],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

View File

@ -0,0 +1,111 @@
/*
Copyright 2018 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 v1alpha1
import (
"net"
"strconv"
"k8s.io/apimachinery/pkg/runtime"
apiserverconfigv1alpha1 "k8s.io/apiserver/pkg/apis/config/v1alpha1"
kubescedulerconfigv1alpha1 "k8s.io/kube-scheduler/config/v1alpha1"
// this package shouldn't really depend on other k8s.io/kubernetes code
api "k8s.io/kubernetes/pkg/apis/core"
kubeletapis "k8s.io/kubernetes/pkg/kubelet/apis"
"k8s.io/kubernetes/pkg/master/ports"
)
func addDefaultingFuncs(scheme *runtime.Scheme) error {
return RegisterDefaults(scheme)
}
// SetDefaults_KubeSchedulerConfiguration sets additional defaults
func SetDefaults_KubeSchedulerConfiguration(obj *kubescedulerconfigv1alpha1.KubeSchedulerConfiguration) {
if len(obj.SchedulerName) == 0 {
obj.SchedulerName = api.DefaultSchedulerName
}
if obj.HardPodAffinitySymmetricWeight == 0 {
obj.HardPodAffinitySymmetricWeight = api.DefaultHardPodAffinitySymmetricWeight
}
if obj.AlgorithmSource.Policy == nil &&
(obj.AlgorithmSource.Provider == nil || len(*obj.AlgorithmSource.Provider) == 0) {
val := kubescedulerconfigv1alpha1.SchedulerDefaultProviderName
obj.AlgorithmSource.Provider = &val
}
if policy := obj.AlgorithmSource.Policy; policy != nil {
if policy.ConfigMap != nil && len(policy.ConfigMap.Namespace) == 0 {
obj.AlgorithmSource.Policy.ConfigMap.Namespace = api.NamespaceSystem
}
}
if host, port, err := net.SplitHostPort(obj.HealthzBindAddress); err == nil {
if len(host) == 0 {
host = "0.0.0.0"
}
obj.HealthzBindAddress = net.JoinHostPort(host, port)
} else {
obj.HealthzBindAddress = net.JoinHostPort("0.0.0.0", strconv.Itoa(ports.InsecureSchedulerPort))
}
if host, port, err := net.SplitHostPort(obj.MetricsBindAddress); err == nil {
if len(host) == 0 {
host = "0.0.0.0"
}
obj.MetricsBindAddress = net.JoinHostPort(host, port)
} else {
obj.MetricsBindAddress = net.JoinHostPort("0.0.0.0", strconv.Itoa(ports.InsecureSchedulerPort))
}
if len(obj.LeaderElection.LockObjectNamespace) == 0 {
obj.LeaderElection.LockObjectNamespace = kubescedulerconfigv1alpha1.SchedulerDefaultLockObjectNamespace
}
if len(obj.LeaderElection.LockObjectName) == 0 {
obj.LeaderElection.LockObjectName = kubescedulerconfigv1alpha1.SchedulerDefaultLockObjectName
}
if obj.PercentageOfNodesToScore == 0 {
// by default, stop finding feasible nodes once the number of feasible nodes is 50% of the cluster.
obj.PercentageOfNodesToScore = 50
}
if len(obj.FailureDomains) == 0 {
obj.FailureDomains = kubeletapis.DefaultFailureDomains
}
if len(obj.ClientConnection.ContentType) == 0 {
obj.ClientConnection.ContentType = "application/vnd.kubernetes.protobuf"
}
// Scheduler has an opinion about QPS/Burst, setting specific defaults for itself, instead of generic settings.
if obj.ClientConnection.QPS == 0.0 {
obj.ClientConnection.QPS = 50.0
}
if obj.ClientConnection.Burst == 0 {
obj.ClientConnection.Burst = 100
}
// Use the default LeaderElectionConfiguration options
apiserverconfigv1alpha1.RecommendedDefaultLeaderElectionConfiguration(&obj.LeaderElection.LeaderElectionConfiguration)
if obj.BindTimeoutSeconds == nil {
defaultBindTimeoutSeconds := int64(600)
obj.BindTimeoutSeconds = &defaultBindTimeoutSeconds
}
}

View File

@ -0,0 +1,64 @@
/*
Copyright 2018 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 v1alpha1
import (
"encoding/json"
"reflect"
"testing"
"k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/runtime"
kubeschedulerconfigv1alpha1 "k8s.io/kube-scheduler/config/v1alpha1"
)
func TestSchedulerDefaults(t *testing.T) {
ks1 := &kubeschedulerconfigv1alpha1.KubeSchedulerConfiguration{}
SetDefaults_KubeSchedulerConfiguration(ks1)
cm, err := convertObjToConfigMap("KubeSchedulerConfiguration", ks1)
if err != nil {
t.Errorf("unexpected ConvertObjToConfigMap error %v", err)
}
ks2 := &kubeschedulerconfigv1alpha1.KubeSchedulerConfiguration{}
if err = json.Unmarshal([]byte(cm.Data["KubeSchedulerConfiguration"]), ks2); err != nil {
t.Errorf("unexpected error unserializing scheduler config %v", err)
}
if !reflect.DeepEqual(ks2, ks1) {
t.Errorf("Expected:\n%#v\n\nGot:\n%#v", ks1, ks2)
}
}
// ConvertObjToConfigMap converts an object to a ConfigMap.
// This is specifically meant for ComponentConfigs.
func convertObjToConfigMap(name string, obj runtime.Object) (*v1.ConfigMap, error) {
eJSONBytes, err := json.Marshal(obj)
if err != nil {
return nil, err
}
cm := &v1.ConfigMap{
ObjectMeta: metav1.ObjectMeta{
Name: name,
},
Data: map[string]string{
name: string(eJSONBytes[:]),
},
}
return cm, nil
}

View File

@ -0,0 +1,24 @@
/*
Copyright 2018 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.
*/
// +k8s:deepcopy-gen=package
// +k8s:conversion-gen=k8s.io/kubernetes/pkg/scheduler/apis/config
// +k8s:conversion-gen-external-types=k8s.io/kube-scheduler/config/v1alpha1
// +k8s:defaulter-gen=TypeMeta
// +k8s:defaulter-gen-input=../../../../../vendor/k8s.io/kube-scheduler/config/v1alpha1
// +groupName=kubescheduler.config.k8s.io
package v1alpha1 // import "k8s.io/kubernetes/pkg/scheduler/apis/config/v1alpha1"

View File

@ -0,0 +1,43 @@
/*
Copyright 2018 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 v1alpha1
import (
"k8s.io/apimachinery/pkg/runtime/schema"
kubeschedulerconfigv1alpha1 "k8s.io/kube-scheduler/config/v1alpha1"
)
// GroupName is the group name used in this package
const GroupName = "kubescheduler.config.k8s.io"
// SchemeGroupVersion is group version used to register these objects
var SchemeGroupVersion = schema.GroupVersion{Group: GroupName, Version: "v1alpha1"}
var (
// localSchemeBuilder extends the SchemeBuilder instance with the external types. In this package,
// defaulting and conversion init funcs are registered as well.
localSchemeBuilder = &kubeschedulerconfigv1alpha1.SchemeBuilder
// AddToScheme is a global function that registers this API group & version to a scheme
AddToScheme = localSchemeBuilder.AddToScheme
)
func init() {
// We only register manually written functions here. The registration of the
// generated functions takes place in the generated files. The separation
// makes the code compile even when the generated files are missing.
localSchemeBuilder.Register(addDefaultingFuncs)
}

View File

@ -0,0 +1,274 @@
// +build !ignore_autogenerated
/*
Copyright 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.
*/
// Code generated by conversion-gen. DO NOT EDIT.
package v1alpha1
import (
unsafe "unsafe"
configv1alpha1 "k8s.io/apimachinery/pkg/apis/config/v1alpha1"
conversion "k8s.io/apimachinery/pkg/conversion"
runtime "k8s.io/apimachinery/pkg/runtime"
apisconfigv1alpha1 "k8s.io/apiserver/pkg/apis/config/v1alpha1"
v1alpha1 "k8s.io/kube-scheduler/config/v1alpha1"
config "k8s.io/kubernetes/pkg/scheduler/apis/config"
)
func init() {
localSchemeBuilder.Register(RegisterConversions)
}
// RegisterConversions adds conversion functions to the given scheme.
// Public to allow building arbitrary schemes.
func RegisterConversions(s *runtime.Scheme) error {
if err := s.AddGeneratedConversionFunc((*v1alpha1.KubeSchedulerConfiguration)(nil), (*config.KubeSchedulerConfiguration)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_v1alpha1_KubeSchedulerConfiguration_To_config_KubeSchedulerConfiguration(a.(*v1alpha1.KubeSchedulerConfiguration), b.(*config.KubeSchedulerConfiguration), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*config.KubeSchedulerConfiguration)(nil), (*v1alpha1.KubeSchedulerConfiguration)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_config_KubeSchedulerConfiguration_To_v1alpha1_KubeSchedulerConfiguration(a.(*config.KubeSchedulerConfiguration), b.(*v1alpha1.KubeSchedulerConfiguration), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*v1alpha1.KubeSchedulerLeaderElectionConfiguration)(nil), (*config.KubeSchedulerLeaderElectionConfiguration)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_v1alpha1_KubeSchedulerLeaderElectionConfiguration_To_config_KubeSchedulerLeaderElectionConfiguration(a.(*v1alpha1.KubeSchedulerLeaderElectionConfiguration), b.(*config.KubeSchedulerLeaderElectionConfiguration), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*config.KubeSchedulerLeaderElectionConfiguration)(nil), (*v1alpha1.KubeSchedulerLeaderElectionConfiguration)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_config_KubeSchedulerLeaderElectionConfiguration_To_v1alpha1_KubeSchedulerLeaderElectionConfiguration(a.(*config.KubeSchedulerLeaderElectionConfiguration), b.(*v1alpha1.KubeSchedulerLeaderElectionConfiguration), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*v1alpha1.SchedulerAlgorithmSource)(nil), (*config.SchedulerAlgorithmSource)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_v1alpha1_SchedulerAlgorithmSource_To_config_SchedulerAlgorithmSource(a.(*v1alpha1.SchedulerAlgorithmSource), b.(*config.SchedulerAlgorithmSource), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*config.SchedulerAlgorithmSource)(nil), (*v1alpha1.SchedulerAlgorithmSource)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_config_SchedulerAlgorithmSource_To_v1alpha1_SchedulerAlgorithmSource(a.(*config.SchedulerAlgorithmSource), b.(*v1alpha1.SchedulerAlgorithmSource), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*v1alpha1.SchedulerPolicyConfigMapSource)(nil), (*config.SchedulerPolicyConfigMapSource)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_v1alpha1_SchedulerPolicyConfigMapSource_To_config_SchedulerPolicyConfigMapSource(a.(*v1alpha1.SchedulerPolicyConfigMapSource), b.(*config.SchedulerPolicyConfigMapSource), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*config.SchedulerPolicyConfigMapSource)(nil), (*v1alpha1.SchedulerPolicyConfigMapSource)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_config_SchedulerPolicyConfigMapSource_To_v1alpha1_SchedulerPolicyConfigMapSource(a.(*config.SchedulerPolicyConfigMapSource), b.(*v1alpha1.SchedulerPolicyConfigMapSource), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*v1alpha1.SchedulerPolicyFileSource)(nil), (*config.SchedulerPolicyFileSource)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_v1alpha1_SchedulerPolicyFileSource_To_config_SchedulerPolicyFileSource(a.(*v1alpha1.SchedulerPolicyFileSource), b.(*config.SchedulerPolicyFileSource), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*config.SchedulerPolicyFileSource)(nil), (*v1alpha1.SchedulerPolicyFileSource)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_config_SchedulerPolicyFileSource_To_v1alpha1_SchedulerPolicyFileSource(a.(*config.SchedulerPolicyFileSource), b.(*v1alpha1.SchedulerPolicyFileSource), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*v1alpha1.SchedulerPolicySource)(nil), (*config.SchedulerPolicySource)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_v1alpha1_SchedulerPolicySource_To_config_SchedulerPolicySource(a.(*v1alpha1.SchedulerPolicySource), b.(*config.SchedulerPolicySource), scope)
}); err != nil {
return err
}
if err := s.AddGeneratedConversionFunc((*config.SchedulerPolicySource)(nil), (*v1alpha1.SchedulerPolicySource)(nil), func(a, b interface{}, scope conversion.Scope) error {
return Convert_config_SchedulerPolicySource_To_v1alpha1_SchedulerPolicySource(a.(*config.SchedulerPolicySource), b.(*v1alpha1.SchedulerPolicySource), scope)
}); err != nil {
return err
}
return nil
}
func autoConvert_v1alpha1_KubeSchedulerConfiguration_To_config_KubeSchedulerConfiguration(in *v1alpha1.KubeSchedulerConfiguration, out *config.KubeSchedulerConfiguration, s conversion.Scope) error {
out.SchedulerName = in.SchedulerName
if err := Convert_v1alpha1_SchedulerAlgorithmSource_To_config_SchedulerAlgorithmSource(&in.AlgorithmSource, &out.AlgorithmSource, s); err != nil {
return err
}
out.HardPodAffinitySymmetricWeight = in.HardPodAffinitySymmetricWeight
if err := Convert_v1alpha1_KubeSchedulerLeaderElectionConfiguration_To_config_KubeSchedulerLeaderElectionConfiguration(&in.LeaderElection, &out.LeaderElection, s); err != nil {
return err
}
if err := configv1alpha1.Convert_v1alpha1_ClientConnectionConfiguration_To_config_ClientConnectionConfiguration(&in.ClientConnection, &out.ClientConnection, s); err != nil {
return err
}
out.HealthzBindAddress = in.HealthzBindAddress
out.MetricsBindAddress = in.MetricsBindAddress
if err := apisconfigv1alpha1.Convert_v1alpha1_DebuggingConfiguration_To_config_DebuggingConfiguration(&in.DebuggingConfiguration, &out.DebuggingConfiguration, s); err != nil {
return err
}
out.DisablePreemption = in.DisablePreemption
out.PercentageOfNodesToScore = in.PercentageOfNodesToScore
out.FailureDomains = in.FailureDomains
out.BindTimeoutSeconds = (*int64)(unsafe.Pointer(in.BindTimeoutSeconds))
return nil
}
// Convert_v1alpha1_KubeSchedulerConfiguration_To_config_KubeSchedulerConfiguration is an autogenerated conversion function.
func Convert_v1alpha1_KubeSchedulerConfiguration_To_config_KubeSchedulerConfiguration(in *v1alpha1.KubeSchedulerConfiguration, out *config.KubeSchedulerConfiguration, s conversion.Scope) error {
return autoConvert_v1alpha1_KubeSchedulerConfiguration_To_config_KubeSchedulerConfiguration(in, out, s)
}
func autoConvert_config_KubeSchedulerConfiguration_To_v1alpha1_KubeSchedulerConfiguration(in *config.KubeSchedulerConfiguration, out *v1alpha1.KubeSchedulerConfiguration, s conversion.Scope) error {
out.SchedulerName = in.SchedulerName
if err := Convert_config_SchedulerAlgorithmSource_To_v1alpha1_SchedulerAlgorithmSource(&in.AlgorithmSource, &out.AlgorithmSource, s); err != nil {
return err
}
out.HardPodAffinitySymmetricWeight = in.HardPodAffinitySymmetricWeight
if err := Convert_config_KubeSchedulerLeaderElectionConfiguration_To_v1alpha1_KubeSchedulerLeaderElectionConfiguration(&in.LeaderElection, &out.LeaderElection, s); err != nil {
return err
}
if err := configv1alpha1.Convert_config_ClientConnectionConfiguration_To_v1alpha1_ClientConnectionConfiguration(&in.ClientConnection, &out.ClientConnection, s); err != nil {
return err
}
out.HealthzBindAddress = in.HealthzBindAddress
out.MetricsBindAddress = in.MetricsBindAddress
if err := apisconfigv1alpha1.Convert_config_DebuggingConfiguration_To_v1alpha1_DebuggingConfiguration(&in.DebuggingConfiguration, &out.DebuggingConfiguration, s); err != nil {
return err
}
out.DisablePreemption = in.DisablePreemption
out.PercentageOfNodesToScore = in.PercentageOfNodesToScore
out.FailureDomains = in.FailureDomains
out.BindTimeoutSeconds = (*int64)(unsafe.Pointer(in.BindTimeoutSeconds))
return nil
}
// Convert_config_KubeSchedulerConfiguration_To_v1alpha1_KubeSchedulerConfiguration is an autogenerated conversion function.
func Convert_config_KubeSchedulerConfiguration_To_v1alpha1_KubeSchedulerConfiguration(in *config.KubeSchedulerConfiguration, out *v1alpha1.KubeSchedulerConfiguration, s conversion.Scope) error {
return autoConvert_config_KubeSchedulerConfiguration_To_v1alpha1_KubeSchedulerConfiguration(in, out, s)
}
func autoConvert_v1alpha1_KubeSchedulerLeaderElectionConfiguration_To_config_KubeSchedulerLeaderElectionConfiguration(in *v1alpha1.KubeSchedulerLeaderElectionConfiguration, out *config.KubeSchedulerLeaderElectionConfiguration, s conversion.Scope) error {
if err := apisconfigv1alpha1.Convert_v1alpha1_LeaderElectionConfiguration_To_config_LeaderElectionConfiguration(&in.LeaderElectionConfiguration, &out.LeaderElectionConfiguration, s); err != nil {
return err
}
out.LockObjectNamespace = in.LockObjectNamespace
out.LockObjectName = in.LockObjectName
return nil
}
// Convert_v1alpha1_KubeSchedulerLeaderElectionConfiguration_To_config_KubeSchedulerLeaderElectionConfiguration is an autogenerated conversion function.
func Convert_v1alpha1_KubeSchedulerLeaderElectionConfiguration_To_config_KubeSchedulerLeaderElectionConfiguration(in *v1alpha1.KubeSchedulerLeaderElectionConfiguration, out *config.KubeSchedulerLeaderElectionConfiguration, s conversion.Scope) error {
return autoConvert_v1alpha1_KubeSchedulerLeaderElectionConfiguration_To_config_KubeSchedulerLeaderElectionConfiguration(in, out, s)
}
func autoConvert_config_KubeSchedulerLeaderElectionConfiguration_To_v1alpha1_KubeSchedulerLeaderElectionConfiguration(in *config.KubeSchedulerLeaderElectionConfiguration, out *v1alpha1.KubeSchedulerLeaderElectionConfiguration, s conversion.Scope) error {
if err := apisconfigv1alpha1.Convert_config_LeaderElectionConfiguration_To_v1alpha1_LeaderElectionConfiguration(&in.LeaderElectionConfiguration, &out.LeaderElectionConfiguration, s); err != nil {
return err
}
out.LockObjectNamespace = in.LockObjectNamespace
out.LockObjectName = in.LockObjectName
return nil
}
// Convert_config_KubeSchedulerLeaderElectionConfiguration_To_v1alpha1_KubeSchedulerLeaderElectionConfiguration is an autogenerated conversion function.
func Convert_config_KubeSchedulerLeaderElectionConfiguration_To_v1alpha1_KubeSchedulerLeaderElectionConfiguration(in *config.KubeSchedulerLeaderElectionConfiguration, out *v1alpha1.KubeSchedulerLeaderElectionConfiguration, s conversion.Scope) error {
return autoConvert_config_KubeSchedulerLeaderElectionConfiguration_To_v1alpha1_KubeSchedulerLeaderElectionConfiguration(in, out, s)
}
func autoConvert_v1alpha1_SchedulerAlgorithmSource_To_config_SchedulerAlgorithmSource(in *v1alpha1.SchedulerAlgorithmSource, out *config.SchedulerAlgorithmSource, s conversion.Scope) error {
out.Policy = (*config.SchedulerPolicySource)(unsafe.Pointer(in.Policy))
out.Provider = (*string)(unsafe.Pointer(in.Provider))
return nil
}
// Convert_v1alpha1_SchedulerAlgorithmSource_To_config_SchedulerAlgorithmSource is an autogenerated conversion function.
func Convert_v1alpha1_SchedulerAlgorithmSource_To_config_SchedulerAlgorithmSource(in *v1alpha1.SchedulerAlgorithmSource, out *config.SchedulerAlgorithmSource, s conversion.Scope) error {
return autoConvert_v1alpha1_SchedulerAlgorithmSource_To_config_SchedulerAlgorithmSource(in, out, s)
}
func autoConvert_config_SchedulerAlgorithmSource_To_v1alpha1_SchedulerAlgorithmSource(in *config.SchedulerAlgorithmSource, out *v1alpha1.SchedulerAlgorithmSource, s conversion.Scope) error {
out.Policy = (*v1alpha1.SchedulerPolicySource)(unsafe.Pointer(in.Policy))
out.Provider = (*string)(unsafe.Pointer(in.Provider))
return nil
}
// Convert_config_SchedulerAlgorithmSource_To_v1alpha1_SchedulerAlgorithmSource is an autogenerated conversion function.
func Convert_config_SchedulerAlgorithmSource_To_v1alpha1_SchedulerAlgorithmSource(in *config.SchedulerAlgorithmSource, out *v1alpha1.SchedulerAlgorithmSource, s conversion.Scope) error {
return autoConvert_config_SchedulerAlgorithmSource_To_v1alpha1_SchedulerAlgorithmSource(in, out, s)
}
func autoConvert_v1alpha1_SchedulerPolicyConfigMapSource_To_config_SchedulerPolicyConfigMapSource(in *v1alpha1.SchedulerPolicyConfigMapSource, out *config.SchedulerPolicyConfigMapSource, s conversion.Scope) error {
out.Namespace = in.Namespace
out.Name = in.Name
return nil
}
// Convert_v1alpha1_SchedulerPolicyConfigMapSource_To_config_SchedulerPolicyConfigMapSource is an autogenerated conversion function.
func Convert_v1alpha1_SchedulerPolicyConfigMapSource_To_config_SchedulerPolicyConfigMapSource(in *v1alpha1.SchedulerPolicyConfigMapSource, out *config.SchedulerPolicyConfigMapSource, s conversion.Scope) error {
return autoConvert_v1alpha1_SchedulerPolicyConfigMapSource_To_config_SchedulerPolicyConfigMapSource(in, out, s)
}
func autoConvert_config_SchedulerPolicyConfigMapSource_To_v1alpha1_SchedulerPolicyConfigMapSource(in *config.SchedulerPolicyConfigMapSource, out *v1alpha1.SchedulerPolicyConfigMapSource, s conversion.Scope) error {
out.Namespace = in.Namespace
out.Name = in.Name
return nil
}
// Convert_config_SchedulerPolicyConfigMapSource_To_v1alpha1_SchedulerPolicyConfigMapSource is an autogenerated conversion function.
func Convert_config_SchedulerPolicyConfigMapSource_To_v1alpha1_SchedulerPolicyConfigMapSource(in *config.SchedulerPolicyConfigMapSource, out *v1alpha1.SchedulerPolicyConfigMapSource, s conversion.Scope) error {
return autoConvert_config_SchedulerPolicyConfigMapSource_To_v1alpha1_SchedulerPolicyConfigMapSource(in, out, s)
}
func autoConvert_v1alpha1_SchedulerPolicyFileSource_To_config_SchedulerPolicyFileSource(in *v1alpha1.SchedulerPolicyFileSource, out *config.SchedulerPolicyFileSource, s conversion.Scope) error {
out.Path = in.Path
return nil
}
// Convert_v1alpha1_SchedulerPolicyFileSource_To_config_SchedulerPolicyFileSource is an autogenerated conversion function.
func Convert_v1alpha1_SchedulerPolicyFileSource_To_config_SchedulerPolicyFileSource(in *v1alpha1.SchedulerPolicyFileSource, out *config.SchedulerPolicyFileSource, s conversion.Scope) error {
return autoConvert_v1alpha1_SchedulerPolicyFileSource_To_config_SchedulerPolicyFileSource(in, out, s)
}
func autoConvert_config_SchedulerPolicyFileSource_To_v1alpha1_SchedulerPolicyFileSource(in *config.SchedulerPolicyFileSource, out *v1alpha1.SchedulerPolicyFileSource, s conversion.Scope) error {
out.Path = in.Path
return nil
}
// Convert_config_SchedulerPolicyFileSource_To_v1alpha1_SchedulerPolicyFileSource is an autogenerated conversion function.
func Convert_config_SchedulerPolicyFileSource_To_v1alpha1_SchedulerPolicyFileSource(in *config.SchedulerPolicyFileSource, out *v1alpha1.SchedulerPolicyFileSource, s conversion.Scope) error {
return autoConvert_config_SchedulerPolicyFileSource_To_v1alpha1_SchedulerPolicyFileSource(in, out, s)
}
func autoConvert_v1alpha1_SchedulerPolicySource_To_config_SchedulerPolicySource(in *v1alpha1.SchedulerPolicySource, out *config.SchedulerPolicySource, s conversion.Scope) error {
out.File = (*config.SchedulerPolicyFileSource)(unsafe.Pointer(in.File))
out.ConfigMap = (*config.SchedulerPolicyConfigMapSource)(unsafe.Pointer(in.ConfigMap))
return nil
}
// Convert_v1alpha1_SchedulerPolicySource_To_config_SchedulerPolicySource is an autogenerated conversion function.
func Convert_v1alpha1_SchedulerPolicySource_To_config_SchedulerPolicySource(in *v1alpha1.SchedulerPolicySource, out *config.SchedulerPolicySource, s conversion.Scope) error {
return autoConvert_v1alpha1_SchedulerPolicySource_To_config_SchedulerPolicySource(in, out, s)
}
func autoConvert_config_SchedulerPolicySource_To_v1alpha1_SchedulerPolicySource(in *config.SchedulerPolicySource, out *v1alpha1.SchedulerPolicySource, s conversion.Scope) error {
out.File = (*v1alpha1.SchedulerPolicyFileSource)(unsafe.Pointer(in.File))
out.ConfigMap = (*v1alpha1.SchedulerPolicyConfigMapSource)(unsafe.Pointer(in.ConfigMap))
return nil
}
// Convert_config_SchedulerPolicySource_To_v1alpha1_SchedulerPolicySource is an autogenerated conversion function.
func Convert_config_SchedulerPolicySource_To_v1alpha1_SchedulerPolicySource(in *config.SchedulerPolicySource, out *v1alpha1.SchedulerPolicySource, s conversion.Scope) error {
return autoConvert_config_SchedulerPolicySource_To_v1alpha1_SchedulerPolicySource(in, out, s)
}

View File

@ -0,0 +1,21 @@
// +build !ignore_autogenerated
/*
Copyright 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.
*/
// Code generated by deepcopy-gen. DO NOT EDIT.
package v1alpha1

View File

@ -0,0 +1,40 @@
// +build !ignore_autogenerated
/*
Copyright 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.
*/
// Code generated by defaulter-gen. DO NOT EDIT.
package v1alpha1
import (
runtime "k8s.io/apimachinery/pkg/runtime"
v1alpha1 "k8s.io/kube-scheduler/config/v1alpha1"
)
// RegisterDefaults adds defaulters functions to the given scheme.
// Public to allow building arbitrary schemes.
// All generated defaulters are covering - they call all nested defaulters.
func RegisterDefaults(scheme *runtime.Scheme) error {
scheme.AddTypeDefaultingFunc(&v1alpha1.KubeSchedulerConfiguration{}, func(obj interface{}) {
SetObjectDefaults_KubeSchedulerConfiguration(obj.(*v1alpha1.KubeSchedulerConfiguration))
})
return nil
}
func SetObjectDefaults_KubeSchedulerConfiguration(in *v1alpha1.KubeSchedulerConfiguration) {
SetDefaults_KubeSchedulerConfiguration(in)
}

View File

@ -0,0 +1,41 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
go_library(
name = "go_default_library",
srcs = ["validation.go"],
importpath = "k8s.io/kubernetes/pkg/scheduler/apis/config/validation",
visibility = ["//visibility:public"],
deps = [
"//pkg/scheduler/apis/config:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/config/validation:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/validation:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/validation/field:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/apis/config/validation:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = ["validation_test.go"],
embed = [":go_default_library"],
deps = [
"//pkg/scheduler/apis/config:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/config:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/apis/config:go_default_library",
],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [":package-srcs"],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

View File

@ -0,0 +1,68 @@
/*
Copyright 2018 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 validation
import (
apimachinery "k8s.io/apimachinery/pkg/apis/config/validation"
"k8s.io/apimachinery/pkg/util/validation"
"k8s.io/apimachinery/pkg/util/validation/field"
apiserver "k8s.io/apiserver/pkg/apis/config/validation"
"k8s.io/kubernetes/pkg/scheduler/apis/config"
)
// ValidateKubeSchedulerConfiguration ensures validation of the KubeSchedulerConfiguration struct
func ValidateKubeSchedulerConfiguration(cc *config.KubeSchedulerConfiguration) field.ErrorList {
allErrs := field.ErrorList{}
allErrs = append(allErrs, apimachinery.ValidateClientConnectionConfiguration(&cc.ClientConnection, field.NewPath("clientConnection"))...)
allErrs = append(allErrs, ValidateKubeSchedulerLeaderElectionConfiguration(&cc.LeaderElection, field.NewPath("leaderElection"))...)
if len(cc.SchedulerName) == 0 {
allErrs = append(allErrs, field.Required(field.NewPath("schedulerName"), ""))
}
for _, msg := range validation.IsValidSocketAddr(cc.HealthzBindAddress) {
allErrs = append(allErrs, field.Invalid(field.NewPath("healthzBindAddress"), cc.HealthzBindAddress, msg))
}
for _, msg := range validation.IsValidSocketAddr(cc.MetricsBindAddress) {
allErrs = append(allErrs, field.Invalid(field.NewPath("metricsBindAddress"), cc.MetricsBindAddress, msg))
}
if cc.HardPodAffinitySymmetricWeight < 0 || cc.HardPodAffinitySymmetricWeight > 100 {
allErrs = append(allErrs, field.Invalid(field.NewPath("hardPodAffinitySymmetricWeight"), cc.HardPodAffinitySymmetricWeight, "not in valid range 0-100"))
}
if cc.BindTimeoutSeconds == nil {
allErrs = append(allErrs, field.Required(field.NewPath("bindTimeoutSeconds"), ""))
}
if cc.PercentageOfNodesToScore < 0 || cc.PercentageOfNodesToScore > 100 {
allErrs = append(allErrs, field.Invalid(field.NewPath("percentageOfNodesToScore"),
cc.PercentageOfNodesToScore, "not in valid range 0-100"))
}
return allErrs
}
// ValidateKubeSchedulerLeaderElectionConfiguration ensures validation of the KubeSchedulerLeaderElectionConfiguration struct
func ValidateKubeSchedulerLeaderElectionConfiguration(cc *config.KubeSchedulerLeaderElectionConfiguration, fldPath *field.Path) field.ErrorList {
allErrs := field.ErrorList{}
if !cc.LeaderElectionConfiguration.LeaderElect {
return allErrs
}
allErrs = append(allErrs, apiserver.ValidateLeaderElectionConfiguration(&cc.LeaderElectionConfiguration, field.NewPath("leaderElectionConfiguration"))...)
if len(cc.LockObjectNamespace) == 0 {
allErrs = append(allErrs, field.Required(fldPath.Child("lockObjectNamespace"), ""))
}
if len(cc.LockObjectName) == 0 {
allErrs = append(allErrs, field.Required(fldPath.Child("lockObjectName"), ""))
}
return allErrs
}

View File

@ -0,0 +1,158 @@
/*
Copyright 2018 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 validation
import (
"testing"
"time"
apimachinery "k8s.io/apimachinery/pkg/apis/config"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
apiserver "k8s.io/apiserver/pkg/apis/config"
"k8s.io/kubernetes/pkg/scheduler/apis/config"
)
func TestValidateKubeSchedulerConfiguration(t *testing.T) {
testTimeout := int64(0)
validConfig := &config.KubeSchedulerConfiguration{
SchedulerName: "me",
HealthzBindAddress: "0.0.0.0:10254",
MetricsBindAddress: "0.0.0.0:10254",
HardPodAffinitySymmetricWeight: 80,
ClientConnection: apimachinery.ClientConnectionConfiguration{
AcceptContentTypes: "application/json",
ContentType: "application/json",
QPS: 10,
Burst: 10,
},
AlgorithmSource: config.SchedulerAlgorithmSource{
Policy: &config.SchedulerPolicySource{
ConfigMap: &config.SchedulerPolicyConfigMapSource{
Namespace: "name",
Name: "name",
},
},
},
LeaderElection: config.KubeSchedulerLeaderElectionConfiguration{
LockObjectNamespace: "name",
LockObjectName: "name",
LeaderElectionConfiguration: apiserver.LeaderElectionConfiguration{
ResourceLock: "configmap",
LeaderElect: true,
LeaseDuration: metav1.Duration{Duration: 30 * time.Second},
RenewDeadline: metav1.Duration{Duration: 15 * time.Second},
RetryPeriod: metav1.Duration{Duration: 5 * time.Second},
},
},
BindTimeoutSeconds: &testTimeout,
PercentageOfNodesToScore: 35,
}
HardPodAffinitySymmetricWeightGt100 := validConfig.DeepCopy()
HardPodAffinitySymmetricWeightGt100.HardPodAffinitySymmetricWeight = 120
HardPodAffinitySymmetricWeightLt0 := validConfig.DeepCopy()
HardPodAffinitySymmetricWeightLt0.HardPodAffinitySymmetricWeight = -1
lockObjectNameNotSet := validConfig.DeepCopy()
lockObjectNameNotSet.LeaderElection.LockObjectName = ""
lockObjectNamespaceNotSet := validConfig.DeepCopy()
lockObjectNamespaceNotSet.LeaderElection.LockObjectNamespace = ""
metricsBindAddrHostInvalid := validConfig.DeepCopy()
metricsBindAddrHostInvalid.MetricsBindAddress = "0.0.0.0.0:9090"
metricsBindAddrPortInvalid := validConfig.DeepCopy()
metricsBindAddrPortInvalid.MetricsBindAddress = "0.0.0.0:909090"
healthzBindAddrHostInvalid := validConfig.DeepCopy()
healthzBindAddrHostInvalid.HealthzBindAddress = "0.0.0.0.0:9090"
healthzBindAddrPortInvalid := validConfig.DeepCopy()
healthzBindAddrPortInvalid.HealthzBindAddress = "0.0.0.0:909090"
enableContentProfilingSetWithoutEnableProfiling := validConfig.DeepCopy()
enableContentProfilingSetWithoutEnableProfiling.EnableProfiling = false
enableContentProfilingSetWithoutEnableProfiling.EnableContentionProfiling = true
bindTimeoutUnset := validConfig.DeepCopy()
bindTimeoutUnset.BindTimeoutSeconds = nil
percentageOfNodesToScore101 := validConfig.DeepCopy()
percentageOfNodesToScore101.PercentageOfNodesToScore = int32(101)
scenarios := map[string]struct {
expectedToFail bool
config *config.KubeSchedulerConfiguration
}{
"good": {
expectedToFail: false,
config: validConfig,
},
"bad-lock-object-names-not-set": {
expectedToFail: true,
config: lockObjectNameNotSet,
},
"bad-lock-object-namespace-not-set": {
expectedToFail: true,
config: lockObjectNamespaceNotSet,
},
"bad-healthz-port-invalid": {
expectedToFail: true,
config: healthzBindAddrPortInvalid,
},
"bad-healthz-host-invalid": {
expectedToFail: true,
config: healthzBindAddrHostInvalid,
},
"bad-metrics-port-invalid": {
expectedToFail: true,
config: metricsBindAddrPortInvalid,
},
"bad-metrics-host-invalid": {
expectedToFail: true,
config: metricsBindAddrHostInvalid,
},
"bad-hard-pod-affinity-symmetric-weight-lt-0": {
expectedToFail: true,
config: HardPodAffinitySymmetricWeightGt100,
},
"bad-hard-pod-affinity-symmetric-weight-gt-100": {
expectedToFail: true,
config: HardPodAffinitySymmetricWeightLt0,
},
"bind-timeout-unset": {
expectedToFail: true,
config: bindTimeoutUnset,
},
"bad-percentage-of-nodes-to-score": {
expectedToFail: true,
config: percentageOfNodesToScore101,
},
}
for name, scenario := range scenarios {
errs := ValidateKubeSchedulerConfiguration(scenario.config)
if len(errs) == 0 && scenario.expectedToFail {
t.Errorf("Unexpected success for scenario: %s", name)
}
if len(errs) > 0 && !scenario.expectedToFail {
t.Errorf("Unexpected failure for scenario: %s - %+v", name, errs)
}
}
}

View File

@ -0,0 +1,160 @@
// +build !ignore_autogenerated
/*
Copyright 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.
*/
// Code generated by deepcopy-gen. DO NOT EDIT.
package config
import (
runtime "k8s.io/apimachinery/pkg/runtime"
)
// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
func (in *KubeSchedulerConfiguration) DeepCopyInto(out *KubeSchedulerConfiguration) {
*out = *in
out.TypeMeta = in.TypeMeta
in.AlgorithmSource.DeepCopyInto(&out.AlgorithmSource)
out.LeaderElection = in.LeaderElection
out.ClientConnection = in.ClientConnection
out.DebuggingConfiguration = in.DebuggingConfiguration
if in.BindTimeoutSeconds != nil {
in, out := &in.BindTimeoutSeconds, &out.BindTimeoutSeconds
*out = new(int64)
**out = **in
}
return
}
// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new KubeSchedulerConfiguration.
func (in *KubeSchedulerConfiguration) DeepCopy() *KubeSchedulerConfiguration {
if in == nil {
return nil
}
out := new(KubeSchedulerConfiguration)
in.DeepCopyInto(out)
return out
}
// DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.
func (in *KubeSchedulerConfiguration) DeepCopyObject() runtime.Object {
if c := in.DeepCopy(); c != nil {
return c
}
return nil
}
// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
func (in *KubeSchedulerLeaderElectionConfiguration) DeepCopyInto(out *KubeSchedulerLeaderElectionConfiguration) {
*out = *in
out.LeaderElectionConfiguration = in.LeaderElectionConfiguration
return
}
// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new KubeSchedulerLeaderElectionConfiguration.
func (in *KubeSchedulerLeaderElectionConfiguration) DeepCopy() *KubeSchedulerLeaderElectionConfiguration {
if in == nil {
return nil
}
out := new(KubeSchedulerLeaderElectionConfiguration)
in.DeepCopyInto(out)
return out
}
// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
func (in *SchedulerAlgorithmSource) DeepCopyInto(out *SchedulerAlgorithmSource) {
*out = *in
if in.Policy != nil {
in, out := &in.Policy, &out.Policy
*out = new(SchedulerPolicySource)
(*in).DeepCopyInto(*out)
}
if in.Provider != nil {
in, out := &in.Provider, &out.Provider
*out = new(string)
**out = **in
}
return
}
// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new SchedulerAlgorithmSource.
func (in *SchedulerAlgorithmSource) DeepCopy() *SchedulerAlgorithmSource {
if in == nil {
return nil
}
out := new(SchedulerAlgorithmSource)
in.DeepCopyInto(out)
return out
}
// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
func (in *SchedulerPolicyConfigMapSource) DeepCopyInto(out *SchedulerPolicyConfigMapSource) {
*out = *in
return
}
// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new SchedulerPolicyConfigMapSource.
func (in *SchedulerPolicyConfigMapSource) DeepCopy() *SchedulerPolicyConfigMapSource {
if in == nil {
return nil
}
out := new(SchedulerPolicyConfigMapSource)
in.DeepCopyInto(out)
return out
}
// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
func (in *SchedulerPolicyFileSource) DeepCopyInto(out *SchedulerPolicyFileSource) {
*out = *in
return
}
// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new SchedulerPolicyFileSource.
func (in *SchedulerPolicyFileSource) DeepCopy() *SchedulerPolicyFileSource {
if in == nil {
return nil
}
out := new(SchedulerPolicyFileSource)
in.DeepCopyInto(out)
return out
}
// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
func (in *SchedulerPolicySource) DeepCopyInto(out *SchedulerPolicySource) {
*out = *in
if in.File != nil {
in, out := &in.File, &out.File
*out = new(SchedulerPolicyFileSource)
**out = **in
}
if in.ConfigMap != nil {
in, out := &in.ConfigMap, &out.ConfigMap
*out = new(SchedulerPolicyConfigMapSource)
**out = **in
}
return
}
// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new SchedulerPolicySource.
func (in *SchedulerPolicySource) DeepCopy() *SchedulerPolicySource {
if in == nil {
return nil
}
out := new(SchedulerPolicySource)
in.DeepCopyInto(out)
return out
}

View File

@ -3,8 +3,7 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
go_library(
name = "go_default_library",
srcs = [
"cache.go",
"interface.go",
"host_ports.go",
"node_info.go",
"util.go",
],
@ -12,40 +11,28 @@ go_library(
visibility = ["//visibility:public"],
deps = [
"//pkg/apis/core/v1/helper:go_default_library",
"//pkg/features:go_default_library",
"//pkg/scheduler/algorithm/priorities/util:go_default_library",
"//pkg/scheduler/util:go_default_library",
"//vendor/github.com/golang/glog:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/policy/v1beta1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/wait:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = [
"cache_test.go",
"host_ports_test.go",
"node_info_test.go",
"util_test.go",
],
embed = [":go_default_library"],
deps = [
"//pkg/features:go_default_library",
"//pkg/scheduler/algorithm/priorities/util:go_default_library",
"//pkg/scheduler/util:go_default_library",
"//pkg/util/parsers:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/policy/v1beta1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/types:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/intstr:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/wait:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/types:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
],
)

View File

@ -0,0 +1,135 @@
/*
Copyright 2018 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 cache
import (
"k8s.io/api/core/v1"
)
// DefaultBindAllHostIP defines the default ip address used to bind to all host.
const DefaultBindAllHostIP = "0.0.0.0"
// ProtocolPort represents a protocol port pair, e.g. tcp:80.
type ProtocolPort struct {
Protocol string
Port int32
}
// NewProtocolPort creates a ProtocolPort instance.
func NewProtocolPort(protocol string, port int32) *ProtocolPort {
pp := &ProtocolPort{
Protocol: protocol,
Port: port,
}
if len(pp.Protocol) == 0 {
pp.Protocol = string(v1.ProtocolTCP)
}
return pp
}
// HostPortInfo stores mapping from ip to a set of ProtocolPort
type HostPortInfo map[string]map[ProtocolPort]struct{}
// Add adds (ip, protocol, port) to HostPortInfo
func (h HostPortInfo) Add(ip, protocol string, port int32) {
if port <= 0 {
return
}
h.sanitize(&ip, &protocol)
pp := NewProtocolPort(protocol, port)
if _, ok := h[ip]; !ok {
h[ip] = map[ProtocolPort]struct{}{
*pp: {},
}
return
}
h[ip][*pp] = struct{}{}
}
// Remove removes (ip, protocol, port) from HostPortInfo
func (h HostPortInfo) Remove(ip, protocol string, port int32) {
if port <= 0 {
return
}
h.sanitize(&ip, &protocol)
pp := NewProtocolPort(protocol, port)
if m, ok := h[ip]; ok {
delete(m, *pp)
if len(h[ip]) == 0 {
delete(h, ip)
}
}
}
// Len returns the total number of (ip, protocol, port) tuple in HostPortInfo
func (h HostPortInfo) Len() int {
length := 0
for _, m := range h {
length += len(m)
}
return length
}
// CheckConflict checks if the input (ip, protocol, port) conflicts with the existing
// ones in HostPortInfo.
func (h HostPortInfo) CheckConflict(ip, protocol string, port int32) bool {
if port <= 0 {
return false
}
h.sanitize(&ip, &protocol)
pp := NewProtocolPort(protocol, port)
// If ip is 0.0.0.0 check all IP's (protocol, port) pair
if ip == DefaultBindAllHostIP {
for _, m := range h {
if _, ok := m[*pp]; ok {
return true
}
}
return false
}
// If ip isn't 0.0.0.0, only check IP and 0.0.0.0's (protocol, port) pair
for _, key := range []string{DefaultBindAllHostIP, ip} {
if m, ok := h[key]; ok {
if _, ok2 := m[*pp]; ok2 {
return true
}
}
}
return false
}
// sanitize the parameters
func (h HostPortInfo) sanitize(ip, protocol *string) {
if len(*ip) == 0 {
*ip = DefaultBindAllHostIP
}
if len(*protocol) == 0 {
*protocol = string(v1.ProtocolTCP)
}
}

View File

@ -0,0 +1,231 @@
/*
Copyright 2018 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 cache
import (
"testing"
)
type hostPortInfoParam struct {
protocol, ip string
port int32
}
func TestHostPortInfo_AddRemove(t *testing.T) {
tests := []struct {
desc string
added []hostPortInfoParam
removed []hostPortInfoParam
length int
}{
{
desc: "normal add case",
added: []hostPortInfoParam{
{"TCP", "127.0.0.1", 79},
{"UDP", "127.0.0.1", 80},
{"TCP", "127.0.0.1", 81},
{"TCP", "127.0.0.1", 82},
// this might not make sense in real case, but the struct doesn't forbid it.
{"TCP", "0.0.0.0", 79},
{"UDP", "0.0.0.0", 80},
{"TCP", "0.0.0.0", 81},
{"TCP", "0.0.0.0", 82},
{"TCP", "0.0.0.0", 0},
{"TCP", "0.0.0.0", -1},
},
length: 8,
},
{
desc: "empty ip and protocol add should work",
added: []hostPortInfoParam{
{"", "127.0.0.1", 79},
{"UDP", "127.0.0.1", 80},
{"", "127.0.0.1", 81},
{"", "127.0.0.1", 82},
{"", "", 79},
{"UDP", "", 80},
{"", "", 81},
{"", "", 82},
{"", "", 0},
{"", "", -1},
},
length: 8,
},
{
desc: "normal remove case",
added: []hostPortInfoParam{
{"TCP", "127.0.0.1", 79},
{"UDP", "127.0.0.1", 80},
{"TCP", "127.0.0.1", 81},
{"TCP", "127.0.0.1", 82},
{"TCP", "0.0.0.0", 79},
{"UDP", "0.0.0.0", 80},
{"TCP", "0.0.0.0", 81},
{"TCP", "0.0.0.0", 82},
},
removed: []hostPortInfoParam{
{"TCP", "127.0.0.1", 79},
{"UDP", "127.0.0.1", 80},
{"TCP", "127.0.0.1", 81},
{"TCP", "127.0.0.1", 82},
{"TCP", "0.0.0.0", 79},
{"UDP", "0.0.0.0", 80},
{"TCP", "0.0.0.0", 81},
{"TCP", "0.0.0.0", 82},
},
length: 0,
},
{
desc: "empty ip and protocol remove should work",
added: []hostPortInfoParam{
{"TCP", "127.0.0.1", 79},
{"UDP", "127.0.0.1", 80},
{"TCP", "127.0.0.1", 81},
{"TCP", "127.0.0.1", 82},
{"TCP", "0.0.0.0", 79},
{"UDP", "0.0.0.0", 80},
{"TCP", "0.0.0.0", 81},
{"TCP", "0.0.0.0", 82},
},
removed: []hostPortInfoParam{
{"", "127.0.0.1", 79},
{"", "127.0.0.1", 81},
{"", "127.0.0.1", 82},
{"UDP", "127.0.0.1", 80},
{"", "", 79},
{"", "", 81},
{"", "", 82},
{"UDP", "", 80},
},
length: 0,
},
}
for _, test := range tests {
hp := make(HostPortInfo)
for _, param := range test.added {
hp.Add(param.ip, param.protocol, param.port)
}
for _, param := range test.removed {
hp.Remove(param.ip, param.protocol, param.port)
}
if hp.Len() != test.length {
t.Errorf("%v failed: expect length %d; got %d", test.desc, test.length, hp.Len())
t.Error(hp)
}
}
}
func TestHostPortInfo_Check(t *testing.T) {
tests := []struct {
desc string
added []hostPortInfoParam
check hostPortInfoParam
expect bool
}{
{
desc: "empty check should check 0.0.0.0 and TCP",
added: []hostPortInfoParam{
{"TCP", "127.0.0.1", 80},
},
check: hostPortInfoParam{"", "", 81},
expect: false,
},
{
desc: "empty check should check 0.0.0.0 and TCP (conflicted)",
added: []hostPortInfoParam{
{"TCP", "127.0.0.1", 80},
},
check: hostPortInfoParam{"", "", 80},
expect: true,
},
{
desc: "empty port check should pass",
added: []hostPortInfoParam{
{"TCP", "127.0.0.1", 80},
},
check: hostPortInfoParam{"", "", 0},
expect: false,
},
{
desc: "0.0.0.0 should check all registered IPs",
added: []hostPortInfoParam{
{"TCP", "127.0.0.1", 80},
},
check: hostPortInfoParam{"TCP", "0.0.0.0", 80},
expect: true,
},
{
desc: "0.0.0.0 with different protocol should be allowed",
added: []hostPortInfoParam{
{"UDP", "127.0.0.1", 80},
},
check: hostPortInfoParam{"TCP", "0.0.0.0", 80},
expect: false,
},
{
desc: "0.0.0.0 with different port should be allowed",
added: []hostPortInfoParam{
{"TCP", "127.0.0.1", 79},
{"TCP", "127.0.0.1", 81},
{"TCP", "127.0.0.1", 82},
},
check: hostPortInfoParam{"TCP", "0.0.0.0", 80},
expect: false,
},
{
desc: "normal ip should check all registered 0.0.0.0",
added: []hostPortInfoParam{
{"TCP", "0.0.0.0", 80},
},
check: hostPortInfoParam{"TCP", "127.0.0.1", 80},
expect: true,
},
{
desc: "normal ip with different port/protocol should be allowed (0.0.0.0)",
added: []hostPortInfoParam{
{"TCP", "0.0.0.0", 79},
{"UDP", "0.0.0.0", 80},
{"TCP", "0.0.0.0", 81},
{"TCP", "0.0.0.0", 82},
},
check: hostPortInfoParam{"TCP", "127.0.0.1", 80},
expect: false,
},
{
desc: "normal ip with different port/protocol should be allowed",
added: []hostPortInfoParam{
{"TCP", "127.0.0.1", 79},
{"UDP", "127.0.0.1", 80},
{"TCP", "127.0.0.1", 81},
{"TCP", "127.0.0.1", 82},
},
check: hostPortInfoParam{"TCP", "127.0.0.1", 80},
expect: false,
},
}
for _, test := range tests {
hp := make(HostPortInfo)
for _, param := range test.added {
hp.Add(param.ip, param.protocol, param.port)
}
if hp.CheckConflict(test.check.ip, test.check.protocol, test.check.port) != test.expect {
t.Errorf("%v failed, expected %t; got %t", test.desc, test.expect, !test.expect)
}
}
}

View File

@ -22,13 +22,12 @@ import (
"sync"
"sync/atomic"
"github.com/golang/glog"
"k8s.io/klog"
"k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/api/resource"
v1helper "k8s.io/kubernetes/pkg/apis/core/v1/helper"
priorityutil "k8s.io/kubernetes/pkg/scheduler/algorithm/priorities/util"
"k8s.io/kubernetes/pkg/scheduler/util"
)
var (
@ -36,6 +35,14 @@ var (
generation int64
)
// ImageStateSummary provides summarized information about the state of an image.
type ImageStateSummary struct {
// Size of the image
Size int64
// Used to track how many nodes have this image
NumNodes int
}
// NodeInfo is node level aggregated information.
type NodeInfo struct {
// Overall node information.
@ -43,7 +50,7 @@ type NodeInfo struct {
pods []*v1.Pod
podsWithAffinity []*v1.Pod
usedPorts util.HostPortInfo
usedPorts HostPortInfo
// Total requested resource of all pods on this node.
// It includes assumed pods which scheduler sends binding to apiserver but
@ -58,14 +65,15 @@ type NodeInfo struct {
taints []v1.Taint
taintsErr error
// This is a map from image name to image size, also for checking image existence on the node
// Cache it here to avoid rebuilding the map during scheduling, e.g., in image_locality.go
imageSizes map[string]int64
// imageStates holds the entry of an image if and only if this image is on the node. The entry can be used for
// checking an image's existence and advanced usage (e.g., image locality scheduling policy) based on the image
// state information.
imageStates map[string]*ImageStateSummary
// TransientInfo holds the information pertaining to a scheduling cycle. This will be destructed at the end of
// scheduling cycle.
// TODO: @ravig. Remove this once we have a clear approach for message passing across predicates and priorities.
TransientInfo *transientSchedulerInfo
TransientInfo *TransientSchedulerInfo
// Cached conditions of node for faster lookup.
memoryPressureCondition v1.ConditionStatus
@ -98,28 +106,28 @@ type nodeTransientInfo struct {
RequestedVolumes int
}
// transientSchedulerInfo is a transient structure which is destructed at the end of each scheduling cycle.
// TransientSchedulerInfo is a transient structure which is destructed at the end of each scheduling cycle.
// It consists of items that are valid for a scheduling cycle and is used for message passing across predicates and
// priorities. Some examples which could be used as fields are number of volumes being used on node, current utilization
// on node etc.
// IMPORTANT NOTE: Make sure that each field in this structure is documented along with usage. Expand this structure
// only when absolutely needed as this data structure will be created and destroyed during every scheduling cycle.
type transientSchedulerInfo struct {
type TransientSchedulerInfo struct {
TransientLock sync.Mutex
// NodeTransInfo holds the information related to nodeTransientInformation. NodeName is the key here.
TransNodeInfo nodeTransientInfo
}
// newTransientSchedulerInfo returns a new scheduler transient structure with initialized values.
func newTransientSchedulerInfo() *transientSchedulerInfo {
tsi := &transientSchedulerInfo{
// NewTransientSchedulerInfo returns a new scheduler transient structure with initialized values.
func NewTransientSchedulerInfo() *TransientSchedulerInfo {
tsi := &TransientSchedulerInfo{
TransNodeInfo: initializeNodeTransientInfo(),
}
return tsi
}
// resetTransientSchedulerInfo resets the transientSchedulerInfo.
func (transientSchedInfo *transientSchedulerInfo) resetTransientSchedulerInfo() {
// ResetTransientSchedulerInfo resets the TransientSchedulerInfo.
func (transientSchedInfo *TransientSchedulerInfo) ResetTransientSchedulerInfo() {
transientSchedInfo.TransientLock.Lock()
defer transientSchedInfo.TransientLock.Unlock()
// Reset TransientNodeInfo.
@ -258,10 +266,10 @@ func NewNodeInfo(pods ...*v1.Pod) *NodeInfo {
requestedResource: &Resource{},
nonzeroRequest: &Resource{},
allocatableResource: &Resource{},
TransientInfo: newTransientSchedulerInfo(),
TransientInfo: NewTransientSchedulerInfo(),
generation: nextGeneration(),
usedPorts: make(util.HostPortInfo),
imageSizes: make(map[string]int64),
usedPorts: make(HostPortInfo),
imageStates: make(map[string]*ImageStateSummary),
}
for _, pod := range pods {
ni.AddPod(pod)
@ -285,20 +293,35 @@ func (n *NodeInfo) Pods() []*v1.Pod {
return n.pods
}
// SetPods sets all pods scheduled (including assumed to be) on this node.
func (n *NodeInfo) SetPods(pods []*v1.Pod) {
n.pods = pods
}
// UsedPorts returns used ports on this node.
func (n *NodeInfo) UsedPorts() util.HostPortInfo {
func (n *NodeInfo) UsedPorts() HostPortInfo {
if n == nil {
return nil
}
return n.usedPorts
}
// ImageSizes returns the image size information on this node.
func (n *NodeInfo) ImageSizes() map[string]int64 {
// SetUsedPorts sets the used ports on this node.
func (n *NodeInfo) SetUsedPorts(newUsedPorts HostPortInfo) {
n.usedPorts = newUsedPorts
}
// ImageStates returns the state information of all images.
func (n *NodeInfo) ImageStates() map[string]*ImageStateSummary {
if n == nil {
return nil
}
return n.imageSizes
return n.imageStates
}
// SetImageStates sets the state information of all images.
func (n *NodeInfo) SetImageStates(newImageStates map[string]*ImageStateSummary) {
n.imageStates = newImageStates
}
// PodsWithAffinity return all pods with (anti)affinity constraints on this node.
@ -325,6 +348,11 @@ func (n *NodeInfo) Taints() ([]v1.Taint, error) {
return n.taints, n.taintsErr
}
// SetTaints sets the taints list on this node.
func (n *NodeInfo) SetTaints(newTaints []v1.Taint) {
n.taints = newTaints
}
// MemoryPressureCondition returns the memory pressure condition status on this node.
func (n *NodeInfo) MemoryPressureCondition() v1.ConditionStatus {
if n == nil {
@ -357,6 +385,11 @@ func (n *NodeInfo) RequestedResource() Resource {
return *n.requestedResource
}
// SetRequestedResource sets the aggregated resource request of pods on this node.
func (n *NodeInfo) SetRequestedResource(newResource *Resource) {
n.requestedResource = newResource
}
// NonZeroRequest returns aggregated nonzero resource request of pods on this node.
func (n *NodeInfo) NonZeroRequest() Resource {
if n == nil {
@ -365,6 +398,11 @@ func (n *NodeInfo) NonZeroRequest() Resource {
return *n.nonzeroRequest
}
// SetNonZeroRequest sets the aggregated nonzero resource request of pods on this node.
func (n *NodeInfo) SetNonZeroRequest(newResource *Resource) {
n.nonzeroRequest = newResource
}
// AllocatableResource returns allocatable resources on a given node.
func (n *NodeInfo) AllocatableResource() Resource {
if n == nil {
@ -379,6 +417,19 @@ func (n *NodeInfo) SetAllocatableResource(allocatableResource *Resource) {
n.generation = nextGeneration()
}
// GetGeneration returns the generation on this node.
func (n *NodeInfo) GetGeneration() int64 {
if n == nil {
return 0
}
return n.generation
}
// SetGeneration sets the generation on this node. This is for testing only.
func (n *NodeInfo) SetGeneration(newGeneration int64) {
n.generation = newGeneration
}
// Clone returns a copy of this node.
func (n *NodeInfo) Clone() *NodeInfo {
clone := &NodeInfo{
@ -391,16 +442,21 @@ func (n *NodeInfo) Clone() *NodeInfo {
memoryPressureCondition: n.memoryPressureCondition,
diskPressureCondition: n.diskPressureCondition,
pidPressureCondition: n.pidPressureCondition,
usedPorts: make(util.HostPortInfo),
imageSizes: n.imageSizes,
usedPorts: make(HostPortInfo),
imageStates: n.imageStates,
generation: n.generation,
}
if len(n.pods) > 0 {
clone.pods = append([]*v1.Pod(nil), n.pods...)
}
if len(n.usedPorts) > 0 {
for k, v := range n.usedPorts {
clone.usedPorts[k] = v
// HostPortInfo is a map-in-map struct
// make sure it's deep copied
for ip, portMap := range n.usedPorts {
clone.usedPorts[ip] = make(map[ProtocolPort]struct{})
for protocolPort, v := range portMap {
clone.usedPorts[ip][protocolPort] = v
}
}
}
if len(n.podsWithAffinity) > 0 {
@ -458,22 +514,22 @@ func (n *NodeInfo) AddPod(pod *v1.Pod) {
}
// Consume ports when pods added.
n.updateUsedPorts(pod, true)
n.UpdateUsedPorts(pod, true)
n.generation = nextGeneration()
}
// RemovePod subtracts pod information from this NodeInfo.
func (n *NodeInfo) RemovePod(pod *v1.Pod) error {
k1, err := getPodKey(pod)
k1, err := GetPodKey(pod)
if err != nil {
return err
}
for i := range n.podsWithAffinity {
k2, err := getPodKey(n.podsWithAffinity[i])
k2, err := GetPodKey(n.podsWithAffinity[i])
if err != nil {
glog.Errorf("Cannot get pod key, err: %v", err)
klog.Errorf("Cannot get pod key, err: %v", err)
continue
}
if k1 == k2 {
@ -484,9 +540,9 @@ func (n *NodeInfo) RemovePod(pod *v1.Pod) error {
}
}
for i := range n.pods {
k2, err := getPodKey(n.pods[i])
k2, err := GetPodKey(n.pods[i])
if err != nil {
glog.Errorf("Cannot get pod key, err: %v", err)
klog.Errorf("Cannot get pod key, err: %v", err)
continue
}
if k1 == k2 {
@ -509,7 +565,7 @@ func (n *NodeInfo) RemovePod(pod *v1.Pod) error {
n.nonzeroRequest.Memory -= non0Mem
// Release ports when remove Pods.
n.updateUsedPorts(pod, false)
n.UpdateUsedPorts(pod, false)
n.generation = nextGeneration()
@ -533,7 +589,8 @@ func calculateResource(pod *v1.Pod) (res Resource, non0CPU int64, non0Mem int64)
return
}
func (n *NodeInfo) updateUsedPorts(pod *v1.Pod, add bool) {
// UpdateUsedPorts updates the UsedPorts of NodeInfo.
func (n *NodeInfo) UpdateUsedPorts(pod *v1.Pod, add bool) {
for j := range pod.Spec.Containers {
container := &pod.Spec.Containers[j]
for k := range container.Ports {
@ -547,17 +604,6 @@ func (n *NodeInfo) updateUsedPorts(pod *v1.Pod, add bool) {
}
}
func (n *NodeInfo) updateImageSizes() {
node := n.Node()
imageSizes := make(map[string]int64)
for _, image := range node.Status.Images {
for _, name := range image.Names {
imageSizes[name] = image.SizeBytes
}
}
n.imageSizes = imageSizes
}
// SetNode sets the overall node information.
func (n *NodeInfo) SetNode(node *v1.Node) error {
n.node = node
@ -578,8 +624,7 @@ func (n *NodeInfo) SetNode(node *v1.Node) error {
// We ignore other conditions.
}
}
n.TransientInfo = newTransientSchedulerInfo()
n.updateImageSizes()
n.TransientInfo = NewTransientSchedulerInfo()
n.generation = nextGeneration()
return nil
}
@ -596,6 +641,7 @@ func (n *NodeInfo) RemoveNode(node *v1.Node) error {
n.memoryPressureCondition = v1.ConditionUnknown
n.diskPressureCondition = v1.ConditionUnknown
n.pidPressureCondition = v1.ConditionUnknown
n.imageStates = make(map[string]*ImageStateSummary)
n.generation = nextGeneration()
return nil
}
@ -619,9 +665,9 @@ func (n *NodeInfo) FilterOutPods(pods []*v1.Pod) []*v1.Pod {
continue
}
// If pod is on the given node, add it to 'filtered' only if it is present in nodeInfo.
podKey, _ := getPodKey(p)
podKey, _ := GetPodKey(p)
for _, np := range n.Pods() {
npodkey, _ := getPodKey(np)
npodkey, _ := GetPodKey(np)
if npodkey == podKey {
filtered = append(filtered, p)
break
@ -631,8 +677,8 @@ func (n *NodeInfo) FilterOutPods(pods []*v1.Pod) []*v1.Pod {
return filtered
}
// getPodKey returns the string key of a pod.
func getPodKey(pod *v1.Pod) (string, error) {
// GetPodKey returns the string key of a pod.
func GetPodKey(pod *v1.Pod) (string, error) {
uid := string(pod.UID)
if len(uid) == 0 {
return "", errors.New("Cannot get cache key for pod with empty UID")

View File

@ -19,14 +19,13 @@ package cache
import (
"fmt"
"reflect"
"strings"
"testing"
"k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/api/resource"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/types"
"k8s.io/kubernetes/pkg/scheduler/util"
"k8s.io/kubernetes/pkg/util/parsers"
)
func TestNewResource(t *testing.T) {
@ -241,44 +240,41 @@ func TestSetMaxResource(t *testing.T) {
}
}
func TestImageSizes(t *testing.T) {
ni := fakeNodeInfo()
ni.node = &v1.Node{
type testingMode interface {
Fatalf(format string, args ...interface{})
}
func makeBasePod(t testingMode, nodeName, objName, cpu, mem, extended string, ports []v1.ContainerPort) *v1.Pod {
req := v1.ResourceList{}
if cpu != "" {
req = v1.ResourceList{
v1.ResourceCPU: resource.MustParse(cpu),
v1.ResourceMemory: resource.MustParse(mem),
}
if extended != "" {
parts := strings.Split(extended, ":")
if len(parts) != 2 {
t.Fatalf("Invalid extended resource string: \"%s\"", extended)
}
req[v1.ResourceName(parts[0])] = resource.MustParse(parts[1])
}
}
return &v1.Pod{
ObjectMeta: metav1.ObjectMeta{
Name: "test-node",
UID: types.UID(objName),
Namespace: "node_info_cache_test",
Name: objName,
},
Status: v1.NodeStatus{
Images: []v1.ContainerImage{
{
Names: []string{
"gcr.io/10:" + parsers.DefaultImageTag,
"gcr.io/10:v1",
},
SizeBytes: int64(10 * 1024 * 1024),
Spec: v1.PodSpec{
Containers: []v1.Container{{
Resources: v1.ResourceRequirements{
Requests: req,
},
{
Names: []string{
"gcr.io/50:" + parsers.DefaultImageTag,
"gcr.io/50:v1",
},
SizeBytes: int64(50 * 1024 * 1024),
},
},
Ports: ports,
}},
NodeName: nodeName,
},
}
ni.updateImageSizes()
expected := map[string]int64{
"gcr.io/10:" + parsers.DefaultImageTag: 10 * 1024 * 1024,
"gcr.io/10:v1": 10 * 1024 * 1024,
"gcr.io/50:" + parsers.DefaultImageTag: 50 * 1024 * 1024,
"gcr.io/50:v1": 50 * 1024 * 1024,
}
imageSizes := ni.ImageSizes()
if !reflect.DeepEqual(expected, imageSizes) {
t.Errorf("expected: %#v, got: %#v", expected, imageSizes)
}
}
func TestNewNodeInfo(t *testing.T) {
@ -303,16 +299,16 @@ func TestNewNodeInfo(t *testing.T) {
AllowedPodNumber: 0,
ScalarResources: map[v1.ResourceName]int64(nil),
},
TransientInfo: newTransientSchedulerInfo(),
TransientInfo: NewTransientSchedulerInfo(),
allocatableResource: &Resource{},
generation: 2,
usedPorts: util.HostPortInfo{
"127.0.0.1": map[util.ProtocolPort]struct{}{
usedPorts: HostPortInfo{
"127.0.0.1": map[ProtocolPort]struct{}{
{Protocol: "TCP", Port: 80}: {},
{Protocol: "TCP", Port: 8080}: {},
},
},
imageSizes: map[string]int64{},
imageStates: map[string]*ImageStateSummary{},
pods: []*v1.Pod{
{
ObjectMeta: metav1.ObjectMeta{
@ -392,18 +388,16 @@ func TestNodeInfoClone(t *testing.T) {
nodeInfo: &NodeInfo{
requestedResource: &Resource{},
nonzeroRequest: &Resource{},
TransientInfo: newTransientSchedulerInfo(),
TransientInfo: NewTransientSchedulerInfo(),
allocatableResource: &Resource{},
generation: 2,
usedPorts: util.HostPortInfo{
"127.0.0.1": map[util.ProtocolPort]struct{}{
usedPorts: HostPortInfo{
"127.0.0.1": map[ProtocolPort]struct{}{
{Protocol: "TCP", Port: 80}: {},
{Protocol: "TCP", Port: 8080}: {},
},
},
imageSizes: map[string]int64{
"gcr.io/10": 10 * 1024 * 1024,
},
imageStates: map[string]*ImageStateSummary{},
pods: []*v1.Pod{
{
ObjectMeta: metav1.ObjectMeta{
@ -464,18 +458,16 @@ func TestNodeInfoClone(t *testing.T) {
expected: &NodeInfo{
requestedResource: &Resource{},
nonzeroRequest: &Resource{},
TransientInfo: newTransientSchedulerInfo(),
TransientInfo: NewTransientSchedulerInfo(),
allocatableResource: &Resource{},
generation: 2,
usedPorts: util.HostPortInfo{
"127.0.0.1": map[util.ProtocolPort]struct{}{
usedPorts: HostPortInfo{
"127.0.0.1": map[ProtocolPort]struct{}{
{Protocol: "TCP", Port: 80}: {},
{Protocol: "TCP", Port: 8080}: {},
},
},
imageSizes: map[string]int64{
"gcr.io/10": 10 * 1024 * 1024,
},
imageStates: map[string]*ImageStateSummary{},
pods: []*v1.Pod{
{
ObjectMeta: metav1.ObjectMeta{
@ -540,6 +532,7 @@ func TestNodeInfoClone(t *testing.T) {
ni := test.nodeInfo.Clone()
// Modify the field to check if the result is a clone of the origin one.
test.nodeInfo.generation += 10
test.nodeInfo.usedPorts.Remove("127.0.0.1", "TCP", 80)
if !reflect.DeepEqual(test.expected, ni) {
t.Errorf("expected: %#v, got: %#v", test.expected, ni)
}
@ -624,16 +617,16 @@ func TestNodeInfoAddPod(t *testing.T) {
AllowedPodNumber: 0,
ScalarResources: map[v1.ResourceName]int64(nil),
},
TransientInfo: newTransientSchedulerInfo(),
TransientInfo: NewTransientSchedulerInfo(),
allocatableResource: &Resource{},
generation: 2,
usedPorts: util.HostPortInfo{
"127.0.0.1": map[util.ProtocolPort]struct{}{
usedPorts: HostPortInfo{
"127.0.0.1": map[ProtocolPort]struct{}{
{Protocol: "TCP", Port: 80}: {},
{Protocol: "TCP", Port: 8080}: {},
},
},
imageSizes: map[string]int64{},
imageStates: map[string]*ImageStateSummary{},
pods: []*v1.Pod{
{
ObjectMeta: metav1.ObjectMeta{
@ -743,16 +736,16 @@ func TestNodeInfoRemovePod(t *testing.T) {
AllowedPodNumber: 0,
ScalarResources: map[v1.ResourceName]int64(nil),
},
TransientInfo: newTransientSchedulerInfo(),
TransientInfo: NewTransientSchedulerInfo(),
allocatableResource: &Resource{},
generation: 2,
usedPorts: util.HostPortInfo{
"127.0.0.1": map[util.ProtocolPort]struct{}{
usedPorts: HostPortInfo{
"127.0.0.1": map[ProtocolPort]struct{}{
{Protocol: "TCP", Port: 80}: {},
{Protocol: "TCP", Port: 8080}: {},
},
},
imageSizes: map[string]int64{},
imageStates: map[string]*ImageStateSummary{},
pods: []*v1.Pod{
{
ObjectMeta: metav1.ObjectMeta{
@ -860,15 +853,15 @@ func TestNodeInfoRemovePod(t *testing.T) {
AllowedPodNumber: 0,
ScalarResources: map[v1.ResourceName]int64(nil),
},
TransientInfo: newTransientSchedulerInfo(),
TransientInfo: NewTransientSchedulerInfo(),
allocatableResource: &Resource{},
generation: 3,
usedPorts: util.HostPortInfo{
"127.0.0.1": map[util.ProtocolPort]struct{}{
usedPorts: HostPortInfo{
"127.0.0.1": map[ProtocolPort]struct{}{
{Protocol: "TCP", Port: 8080}: {},
},
},
imageSizes: map[string]int64{},
imageStates: map[string]*ImageStateSummary{},
pods: []*v1.Pod{
{
ObjectMeta: metav1.ObjectMeta{
@ -909,7 +902,7 @@ func TestNodeInfoRemovePod(t *testing.T) {
err := ni.RemovePod(test.pod)
if err != nil {
if test.errExpected {
expectedErrorMsg := fmt.Errorf("no corresponding pod %s in pods of node %s", test.pod.Name, ni.node.Name)
expectedErrorMsg := fmt.Errorf("no corresponding pod %s in pods of node %s", test.pod.Name, ni.Node().Name)
if expectedErrorMsg == err {
t.Errorf("expected error: %v, got: %v", expectedErrorMsg, err)
}
@ -931,10 +924,10 @@ func TestNodeInfoRemovePod(t *testing.T) {
func fakeNodeInfo(pods ...*v1.Pod) *NodeInfo {
ni := NewNodeInfo(pods...)
ni.node = &v1.Node{
ni.SetNode(&v1.Node{
ObjectMeta: metav1.ObjectMeta{
Name: "test-node",
},
}
})
return ni
}

View File

@ -16,7 +16,10 @@ limitations under the License.
package cache
import "k8s.io/api/core/v1"
import (
"k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/util/sets"
)
// CreateNodeNameToInfoMap obtains a list of pods and pivots that list into a map where the keys are node names
// and the values are the aggregated information for that node.
@ -29,11 +32,47 @@ func CreateNodeNameToInfoMap(pods []*v1.Pod, nodes []*v1.Node) map[string]*NodeI
}
nodeNameToInfo[nodeName].AddPod(pod)
}
imageExistenceMap := createImageExistenceMap(nodes)
for _, node := range nodes {
if _, ok := nodeNameToInfo[node.Name]; !ok {
nodeNameToInfo[node.Name] = NewNodeInfo()
}
nodeNameToInfo[node.Name].SetNode(node)
nodeInfo := nodeNameToInfo[node.Name]
nodeInfo.SetNode(node)
nodeInfo.imageStates = getNodeImageStates(node, imageExistenceMap)
}
return nodeNameToInfo
}
// getNodeImageStates returns the given node's image states based on the given imageExistence map.
func getNodeImageStates(node *v1.Node, imageExistenceMap map[string]sets.String) map[string]*ImageStateSummary {
imageStates := make(map[string]*ImageStateSummary)
for _, image := range node.Status.Images {
for _, name := range image.Names {
imageStates[name] = &ImageStateSummary{
Size: image.SizeBytes,
NumNodes: len(imageExistenceMap[name]),
}
}
}
return imageStates
}
// createImageExistenceMap returns a map recording on which nodes the images exist, keyed by the images' names.
func createImageExistenceMap(nodes []*v1.Node) map[string]sets.String {
imageExistenceMap := make(map[string]sets.String)
for _, node := range nodes {
for _, image := range node.Status.Images {
for _, name := range image.Names {
if _, ok := imageExistenceMap[name]; !ok {
imageExistenceMap[name] = sets.NewString(node.Name)
} else {
imageExistenceMap[name].Insert(node.Name)
}
}
}
}
return imageExistenceMap
}

View File

@ -0,0 +1,134 @@
/*
Copyright 2018 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 cache
import (
"reflect"
"testing"
"k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/util/sets"
)
const mb int64 = 1024 * 1024
func TestGetNodeImageStates(t *testing.T) {
tests := []struct {
node *v1.Node
imageExistenceMap map[string]sets.String
expected map[string]*ImageStateSummary
}{
{
node: &v1.Node{
ObjectMeta: metav1.ObjectMeta{Name: "node-0"},
Status: v1.NodeStatus{
Images: []v1.ContainerImage{
{
Names: []string{
"gcr.io/10:v1",
},
SizeBytes: int64(10 * mb),
},
{
Names: []string{
"gcr.io/200:v1",
},
SizeBytes: int64(200 * mb),
},
},
},
},
imageExistenceMap: map[string]sets.String{
"gcr.io/10:v1": sets.NewString("node-0", "node-1"),
"gcr.io/200:v1": sets.NewString("node-0"),
},
expected: map[string]*ImageStateSummary{
"gcr.io/10:v1": {
Size: int64(10 * mb),
NumNodes: 2,
},
"gcr.io/200:v1": {
Size: int64(200 * mb),
NumNodes: 1,
},
},
},
}
for _, test := range tests {
imageStates := getNodeImageStates(test.node, test.imageExistenceMap)
if !reflect.DeepEqual(test.expected, imageStates) {
t.Errorf("expected: %#v, got: %#v", test.expected, imageStates)
}
}
}
func TestCreateImageExistenceMap(t *testing.T) {
tests := []struct {
nodes []*v1.Node
expected map[string]sets.String
}{
{
nodes: []*v1.Node{
{
ObjectMeta: metav1.ObjectMeta{Name: "node-0"},
Status: v1.NodeStatus{
Images: []v1.ContainerImage{
{
Names: []string{
"gcr.io/10:v1",
},
SizeBytes: int64(10 * mb),
},
},
},
},
{
ObjectMeta: metav1.ObjectMeta{Name: "node-1"},
Status: v1.NodeStatus{
Images: []v1.ContainerImage{
{
Names: []string{
"gcr.io/10:v1",
},
SizeBytes: int64(10 * mb),
},
{
Names: []string{
"gcr.io/200:v1",
},
SizeBytes: int64(200 * mb),
},
},
},
},
},
expected: map[string]sets.String{
"gcr.io/10:v1": sets.NewString("node-0", "node-1"),
"gcr.io/200:v1": sets.NewString("node-1"),
},
},
}
for _, test := range tests {
imageMap := createImageExistenceMap(test.nodes)
if !reflect.DeepEqual(test.expected, imageMap) {
t.Errorf("expected: %#v, got: %#v", test.expected, imageMap)
}
}
}

View File

@ -1,18 +1,44 @@
package(default_visibility = ["//visibility:public"])
load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
load(
"@io_bazel_rules_go//go:def.bzl",
"go_library",
"go_test",
go_library(
name = "go_default_library",
srcs = [
"extender.go",
"generic_scheduler.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/core",
visibility = ["//visibility:public"],
deps = [
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/predicates:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/core/equivalence:go_default_library",
"//pkg/scheduler/internal/cache:go_default_library",
"//pkg/scheduler/internal/queue:go_default_library",
"//pkg/scheduler/metrics:go_default_library",
"//pkg/scheduler/util:go_default_library",
"//pkg/scheduler/volumebinder:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/api/policy/v1beta1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/errors:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/net:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/trace:go_default_library",
"//staging/src/k8s.io/client-go/listers/core/v1:go_default_library",
"//staging/src/k8s.io/client-go/rest:go_default_library",
"//staging/src/k8s.io/client-go/util/workqueue:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = [
"equivalence_cache_test.go",
"extender_test.go",
"generic_scheduler_test.go",
"scheduling_queue_test.go",
],
embed = [":go_default_library"],
deps = [
@ -22,52 +48,19 @@ go_test(
"//pkg/scheduler/algorithm/priorities/util:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/core/equivalence:go_default_library",
"//pkg/scheduler/internal/cache:go_default_library",
"//pkg/scheduler/internal/queue:go_default_library",
"//pkg/scheduler/testing:go_default_library",
"//pkg/scheduler/util:go_default_library",
"//vendor/k8s.io/api/apps/v1beta1:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/extensions/v1beta1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/types:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/wait:go_default_library",
],
)
go_library(
name = "go_default_library",
srcs = [
"equivalence_cache.go",
"extender.go",
"generic_scheduler.go",
"scheduling_queue.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/core",
deps = [
"//pkg/api/v1/pod:go_default_library",
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/predicates:go_default_library",
"//pkg/scheduler/algorithm/priorities/util:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/metrics:go_default_library",
"//pkg/scheduler/util:go_default_library",
"//pkg/scheduler/volumebinder:go_default_library",
"//pkg/util/hash:go_default_library",
"//vendor/github.com/golang/glog:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/policy/v1beta1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/errors:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/net:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/trace:go_default_library",
"//vendor/k8s.io/client-go/listers/core/v1:go_default_library",
"//vendor/k8s.io/client-go/rest:go_default_library",
"//vendor/k8s.io/client-go/tools/cache:go_default_library",
"//vendor/k8s.io/client-go/util/workqueue:go_default_library",
"//staging/src/k8s.io/api/apps/v1:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/types:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/errors:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/wait:go_default_library",
],
)
@ -80,6 +73,10 @@ filegroup(
filegroup(
name = "all-srcs",
srcs = [":package-srcs"],
srcs = [
":package-srcs",
"//pkg/scheduler/core/equivalence:all-srcs",
],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

View File

@ -0,0 +1,49 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
go_library(
name = "go_default_library",
srcs = ["eqivalence.go"],
importpath = "k8s.io/kubernetes/pkg/scheduler/core/equivalence",
visibility = ["//visibility:public"],
deps = [
"//pkg/features:go_default_library",
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/predicates:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/metrics:go_default_library",
"//pkg/util/hash:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = ["eqivalence_test.go"],
embed = [":go_default_library"],
deps = [
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/predicates:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [":package-srcs"],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

View File

@ -0,0 +1,443 @@
/*
Copyright 2016 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 equivalence defines Pod equivalence classes and the equivalence class
// cache.
package equivalence
import (
"fmt"
"hash/fnv"
"sync"
"k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/util/sets"
utilfeature "k8s.io/apiserver/pkg/util/feature"
"k8s.io/klog"
"k8s.io/kubernetes/pkg/features"
"k8s.io/kubernetes/pkg/scheduler/algorithm"
"k8s.io/kubernetes/pkg/scheduler/algorithm/predicates"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
"k8s.io/kubernetes/pkg/scheduler/metrics"
hashutil "k8s.io/kubernetes/pkg/util/hash"
)
// nodeMap stores a *NodeCache for each node.
type nodeMap map[string]*NodeCache
// Cache is a thread safe map saves and reuses the output of predicate functions,
// it uses node name as key to access those cached results.
//
// Internally, results are keyed by predicate name, and "equivalence
// class". (Equivalence class is defined in the `Class` type.) Saved results
// will be reused until an appropriate invalidation function is called.
type Cache struct {
// NOTE(harry): Theoretically sync.Map has better performance in machine with 8+ CPUs, while
// the reality is lock contention in first level cache is rare.
mu sync.RWMutex
nodeToCache nodeMap
predicateIDMap map[string]int
}
// NewCache create an empty equiv class cache.
func NewCache(predicates []string) *Cache {
predicateIDMap := make(map[string]int, len(predicates))
for id, predicate := range predicates {
predicateIDMap[predicate] = id
}
return &Cache{
nodeToCache: make(nodeMap),
predicateIDMap: predicateIDMap,
}
}
// NodeCache saves and reuses the output of predicate functions. Use RunPredicate to
// get or update the cached results. An appropriate Invalidate* function should
// be called when some predicate results are no longer valid.
//
// Internally, results are keyed by predicate name, and "equivalence
// class". (Equivalence class is defined in the `Class` type.) Saved results
// will be reused until an appropriate invalidation function is called.
//
// NodeCache objects are thread safe within the context of NodeCache,
type NodeCache struct {
mu sync.RWMutex
cache predicateMap
// generation is current generation of node cache, incremented on node
// invalidation.
generation uint64
// snapshotGeneration saves snapshot of generation of node cache.
snapshotGeneration uint64
// predicateGenerations stores generation numbers for predicates, incremented on
// predicate invalidation. Created on first update. Use 0 if does not
// exist.
predicateGenerations []uint64
// snapshotPredicateGenerations saves snapshot of generation numbers for predicates.
snapshotPredicateGenerations []uint64
}
// newNodeCache returns an empty NodeCache.
func newNodeCache(n int) *NodeCache {
return &NodeCache{
cache: make(predicateMap, n),
predicateGenerations: make([]uint64, n),
snapshotPredicateGenerations: make([]uint64, n),
}
}
// Snapshot snapshots current generations of cache.
// NOTE: We snapshot generations of all node caches before using it and these
// operations are serialized, we can save snapshot as member of node cache
// itself.
func (c *Cache) Snapshot() {
c.mu.RLock()
defer c.mu.RUnlock()
for _, n := range c.nodeToCache {
n.mu.Lock()
// snapshot predicate generations
copy(n.snapshotPredicateGenerations, n.predicateGenerations)
// snapshot node generation
n.snapshotGeneration = n.generation
n.mu.Unlock()
}
return
}
// GetNodeCache returns the existing NodeCache for given node if present. Otherwise,
// it creates the NodeCache and returns it.
// The boolean flag is true if the value was loaded, false if created.
func (c *Cache) GetNodeCache(name string) (nodeCache *NodeCache, exists bool) {
c.mu.Lock()
defer c.mu.Unlock()
if nodeCache, exists = c.nodeToCache[name]; !exists {
nodeCache = newNodeCache(len(c.predicateIDMap))
c.nodeToCache[name] = nodeCache
}
return
}
// LoadNodeCache returns the existing NodeCache for given node, nil if not
// present.
func (c *Cache) LoadNodeCache(node string) *NodeCache {
c.mu.RLock()
defer c.mu.RUnlock()
return c.nodeToCache[node]
}
func (c *Cache) predicateKeysToIDs(predicateKeys sets.String) []int {
predicateIDs := make([]int, 0, len(predicateKeys))
for predicateKey := range predicateKeys {
if id, ok := c.predicateIDMap[predicateKey]; ok {
predicateIDs = append(predicateIDs, id)
} else {
klog.Errorf("predicate key %q not found", predicateKey)
}
}
return predicateIDs
}
// InvalidatePredicates clears all cached results for the given predicates.
func (c *Cache) InvalidatePredicates(predicateKeys sets.String) {
if len(predicateKeys) == 0 {
return
}
c.mu.RLock()
defer c.mu.RUnlock()
predicateIDs := c.predicateKeysToIDs(predicateKeys)
for _, n := range c.nodeToCache {
n.invalidatePreds(predicateIDs)
}
klog.V(5).Infof("Cache invalidation: node=*,predicates=%v", predicateKeys)
}
// InvalidatePredicatesOnNode clears cached results for the given predicates on one node.
func (c *Cache) InvalidatePredicatesOnNode(nodeName string, predicateKeys sets.String) {
if len(predicateKeys) == 0 {
return
}
c.mu.RLock()
defer c.mu.RUnlock()
predicateIDs := c.predicateKeysToIDs(predicateKeys)
if n, ok := c.nodeToCache[nodeName]; ok {
n.invalidatePreds(predicateIDs)
}
klog.V(5).Infof("Cache invalidation: node=%s,predicates=%v", nodeName, predicateKeys)
}
// InvalidateAllPredicatesOnNode clears all cached results for one node.
func (c *Cache) InvalidateAllPredicatesOnNode(nodeName string) {
c.mu.RLock()
defer c.mu.RUnlock()
if node, ok := c.nodeToCache[nodeName]; ok {
node.invalidate()
}
klog.V(5).Infof("Cache invalidation: node=%s,predicates=*", nodeName)
}
// InvalidateCachedPredicateItemForPodAdd is a wrapper of
// InvalidateCachedPredicateItem for pod add case
// TODO: This does not belong with the equivalence cache implementation.
func (c *Cache) InvalidateCachedPredicateItemForPodAdd(pod *v1.Pod, nodeName string) {
// MatchInterPodAffinity: we assume scheduler can make sure newly bound pod
// will not break the existing inter pod affinity. So we does not need to
// invalidate MatchInterPodAffinity when pod added.
//
// But when a pod is deleted, existing inter pod affinity may become invalid.
// (e.g. this pod was preferred by some else, or vice versa)
//
// NOTE: assumptions above will not stand when we implemented features like
// RequiredDuringSchedulingRequiredDuringExecutioc.
// NoDiskConflict: the newly scheduled pod fits to existing pods on this node,
// it will also fits to equivalence class of existing pods
// GeneralPredicates: will always be affected by adding a new pod
invalidPredicates := sets.NewString(predicates.GeneralPred)
// MaxPDVolumeCountPredicate: we check the volumes of pod to make decisioc.
for _, vol := range pod.Spec.Volumes {
if vol.PersistentVolumeClaim != nil {
invalidPredicates.Insert(
predicates.MaxEBSVolumeCountPred,
predicates.MaxGCEPDVolumeCountPred,
predicates.MaxAzureDiskVolumeCountPred)
if utilfeature.DefaultFeatureGate.Enabled(features.AttachVolumeLimit) {
invalidPredicates.Insert(predicates.MaxCSIVolumeCountPred)
}
} else {
// We do not consider CSI volumes here because CSI
// volumes can not be used inline.
if vol.AWSElasticBlockStore != nil {
invalidPredicates.Insert(predicates.MaxEBSVolumeCountPred)
}
if vol.GCEPersistentDisk != nil {
invalidPredicates.Insert(predicates.MaxGCEPDVolumeCountPred)
}
if vol.AzureDisk != nil {
invalidPredicates.Insert(predicates.MaxAzureDiskVolumeCountPred)
}
}
}
c.InvalidatePredicatesOnNode(nodeName, invalidPredicates)
}
// Class represents a set of pods which are equivalent from the perspective of
// the scheduler. i.e. the scheduler would make the same decision for any pod
// from the same class.
type Class struct {
// Equivalence hash
hash uint64
}
// NewClass returns the equivalence class for a given Pod. The returned Class
// objects will be equal for two Pods in the same class. nil values should not
// be considered equal to each other.
//
// NOTE: Make sure to compare types of Class and not *Class.
// TODO(misterikkit): Return error instead of nil *Class.
func NewClass(pod *v1.Pod) *Class {
equivalencePod := getEquivalencePod(pod)
if equivalencePod != nil {
hash := fnv.New32a()
hashutil.DeepHashObject(hash, equivalencePod)
return &Class{
hash: uint64(hash.Sum32()),
}
}
return nil
}
// predicateMap stores resultMaps with predicate ID as the key.
type predicateMap []resultMap
// resultMap stores PredicateResult with pod equivalence hash as the key.
type resultMap map[uint64]predicateResult
// predicateResult stores the output of a FitPredicate.
type predicateResult struct {
Fit bool
FailReasons []algorithm.PredicateFailureReason
}
// RunPredicate returns a cached predicate result. In case of a cache miss, the predicate will be
// run and its results cached for the next call.
//
// NOTE: RunPredicate will not update the equivalence cache if generation does not match live version.
func (n *NodeCache) RunPredicate(
pred algorithm.FitPredicate,
predicateKey string,
predicateID int,
pod *v1.Pod,
meta algorithm.PredicateMetadata,
nodeInfo *schedulercache.NodeInfo,
equivClass *Class,
) (bool, []algorithm.PredicateFailureReason, error) {
if nodeInfo == nil || nodeInfo.Node() == nil {
// This may happen during tests.
return false, []algorithm.PredicateFailureReason{}, fmt.Errorf("nodeInfo is nil or node is invalid")
}
result, ok := n.lookupResult(pod.GetName(), nodeInfo.Node().GetName(), predicateKey, predicateID, equivClass.hash)
if ok {
return result.Fit, result.FailReasons, nil
}
fit, reasons, err := pred(pod, meta, nodeInfo)
if err != nil {
return fit, reasons, err
}
n.updateResult(pod.GetName(), predicateKey, predicateID, fit, reasons, equivClass.hash, nodeInfo)
return fit, reasons, nil
}
// updateResult updates the cached result of a predicate.
func (n *NodeCache) updateResult(
podName, predicateKey string,
predicateID int,
fit bool,
reasons []algorithm.PredicateFailureReason,
equivalenceHash uint64,
nodeInfo *schedulercache.NodeInfo,
) {
if nodeInfo == nil || nodeInfo.Node() == nil {
// This may happen during tests.
metrics.EquivalenceCacheWrites.WithLabelValues("discarded_bad_node").Inc()
return
}
predicateItem := predicateResult{
Fit: fit,
FailReasons: reasons,
}
n.mu.Lock()
defer n.mu.Unlock()
if (n.snapshotGeneration != n.generation) || (n.snapshotPredicateGenerations[predicateID] != n.predicateGenerations[predicateID]) {
// Generation of node or predicate has been updated since we last took
// a snapshot, this indicates that we received a invalidation request
// during this time. Cache may be stale, skip update.
metrics.EquivalenceCacheWrites.WithLabelValues("discarded_stale").Inc()
return
}
// If cached predicate map already exists, just update the predicate by key
if predicates := n.cache[predicateID]; predicates != nil {
// maps in golang are references, no need to add them back
predicates[equivalenceHash] = predicateItem
} else {
n.cache[predicateID] =
resultMap{
equivalenceHash: predicateItem,
}
}
n.predicateGenerations[predicateID]++
klog.V(5).Infof("Cache update: node=%s, predicate=%s,pod=%s,value=%v",
nodeInfo.Node().Name, predicateKey, podName, predicateItem)
}
// lookupResult returns cached predicate results and a bool saying whether a
// cache entry was found.
func (n *NodeCache) lookupResult(
podName, nodeName, predicateKey string,
predicateID int,
equivalenceHash uint64,
) (value predicateResult, ok bool) {
n.mu.RLock()
defer n.mu.RUnlock()
value, ok = n.cache[predicateID][equivalenceHash]
if ok {
metrics.EquivalenceCacheHits.Inc()
} else {
metrics.EquivalenceCacheMisses.Inc()
}
return value, ok
}
// invalidatePreds deletes cached predicates by given IDs.
func (n *NodeCache) invalidatePreds(predicateIDs []int) {
n.mu.Lock()
defer n.mu.Unlock()
for _, predicateID := range predicateIDs {
n.cache[predicateID] = nil
n.predicateGenerations[predicateID]++
}
}
// invalidate invalidates node cache.
func (n *NodeCache) invalidate() {
n.mu.Lock()
defer n.mu.Unlock()
n.cache = make(predicateMap, len(n.cache))
n.generation++
}
// equivalencePod is the set of pod attributes which must match for two pods to
// be considered equivalent for scheduling purposes. For correctness, this must
// include any Pod field which is used by a FitPredicate.
//
// NOTE: For equivalence hash to be formally correct, lists and maps in the
// equivalencePod should be normalized. (e.g. by sorting them) However, the vast
// majority of equivalent pod classes are expected to be created from a single
// pod template, so they will all have the same ordering.
type equivalencePod struct {
Namespace *string
Labels map[string]string
Affinity *v1.Affinity
Containers []v1.Container // See note about ordering
InitContainers []v1.Container // See note about ordering
NodeName *string
NodeSelector map[string]string
Tolerations []v1.Toleration
Volumes []v1.Volume // See note about ordering
}
// getEquivalencePod returns a normalized representation of a pod so that two
// "equivalent" pods will hash to the same value.
func getEquivalencePod(pod *v1.Pod) *equivalencePod {
ep := &equivalencePod{
Namespace: &pod.Namespace,
Labels: pod.Labels,
Affinity: pod.Spec.Affinity,
Containers: pod.Spec.Containers,
InitContainers: pod.Spec.InitContainers,
NodeName: &pod.Spec.NodeName,
NodeSelector: pod.Spec.NodeSelector,
Tolerations: pod.Spec.Tolerations,
Volumes: pod.Spec.Volumes,
}
// DeepHashObject considers nil and empty slices to be different. Normalize them.
if len(ep.Containers) == 0 {
ep.Containers = nil
}
if len(ep.InitContainers) == 0 {
ep.InitContainers = nil
}
if len(ep.Tolerations) == 0 {
ep.Tolerations = nil
}
if len(ep.Volumes) == 0 {
ep.Volumes = nil
}
// Normalize empty maps also.
if len(ep.Labels) == 0 {
ep.Labels = nil
}
if len(ep.NodeSelector) == 0 {
ep.NodeSelector = nil
}
// TODO(misterikkit): Also normalize nested maps and slices.
return ep
}

View File

@ -14,24 +14,20 @@ See the License for the specific language governing permissions and
limitations under the License.
*/
package core
package equivalence
import (
"errors"
"reflect"
"sync"
"testing"
"time"
"k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/api/resource"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/apimachinery/pkg/util/wait"
"k8s.io/kubernetes/pkg/scheduler/algorithm"
"k8s.io/kubernetes/pkg/scheduler/algorithm/predicates"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
schedulertesting "k8s.io/kubernetes/pkg/scheduler/testing"
)
// makeBasicPod returns a Pod object with many of the fields populated.
@ -158,16 +154,6 @@ type predicateItemType struct {
reasons []algorithm.PredicateFailureReason
}
// upToDateCache is a fake Cache where IsUpToDate always returns true.
type upToDateCache = schedulertesting.FakeCache
// staleNodeCache is a fake Cache where IsUpToDate always returns false.
type staleNodeCache struct {
schedulertesting.FakeCache
}
func (c *staleNodeCache) IsUpToDate(*schedulercache.NodeInfo) bool { return false }
// mockPredicate provides an algorithm.FitPredicate with pre-set return values.
type mockPredicate struct {
fit bool
@ -185,7 +171,6 @@ func TestRunPredicate(t *testing.T) {
tests := []struct {
name string
pred mockPredicate
cache schedulercache.Cache
expectFit, expectCacheHit, expectCacheWrite bool
expectedReasons []algorithm.PredicateFailureReason
expectedError string
@ -193,7 +178,6 @@ func TestRunPredicate(t *testing.T) {
{
name: "pod fits/cache hit",
pred: mockPredicate{},
cache: &upToDateCache{},
expectFit: true,
expectCacheHit: true,
expectCacheWrite: false,
@ -201,23 +185,13 @@ func TestRunPredicate(t *testing.T) {
{
name: "pod fits/cache miss",
pred: mockPredicate{fit: true},
cache: &upToDateCache{},
expectFit: true,
expectCacheHit: false,
expectCacheWrite: true,
},
{
name: "pod fits/cache miss/no write",
pred: mockPredicate{fit: true},
cache: &staleNodeCache{},
expectFit: true,
expectCacheHit: false,
expectCacheWrite: false,
},
{
name: "pod doesn't fit/cache miss",
pred: mockPredicate{reasons: []algorithm.PredicateFailureReason{predicates.ErrFakePredicate}},
cache: &upToDateCache{},
expectFit: false,
expectCacheHit: false,
expectCacheWrite: true,
@ -226,7 +200,6 @@ func TestRunPredicate(t *testing.T) {
{
name: "pod doesn't fit/cache hit",
pred: mockPredicate{},
cache: &upToDateCache{},
expectFit: false,
expectCacheHit: true,
expectCacheWrite: false,
@ -235,7 +208,6 @@ func TestRunPredicate(t *testing.T) {
{
name: "predicate error",
pred: mockPredicate{err: errors.New("This is expected")},
cache: &upToDateCache{},
expectFit: false,
expectCacheHit: false,
expectCacheWrite: false,
@ -243,20 +215,27 @@ func TestRunPredicate(t *testing.T) {
},
}
predicatesOrdering := []string{"testPredicate"}
predicateID := 0
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
node := schedulercache.NewNodeInfo()
node.SetNode(&v1.Node{ObjectMeta: metav1.ObjectMeta{Name: "n1"}})
testNode := &v1.Node{ObjectMeta: metav1.ObjectMeta{Name: "n1"}}
node.SetNode(testNode)
pod := &v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p1"}}
meta := algorithm.EmptyPredicateMetadataProducer(nil, nil)
ecache := NewEquivalenceCache()
equivClass := ecache.getEquivalenceClassInfo(pod)
// Initialize and populate equivalence class cache.
ecache := NewCache(predicatesOrdering)
ecache.Snapshot()
nodeCache, _ := ecache.GetNodeCache(testNode.Name)
equivClass := NewClass(pod)
if test.expectCacheHit {
ecache.updateResult(pod.Name, "testPredicate", test.expectFit, test.expectedReasons, equivClass.hash, test.cache, node)
nodeCache.updateResult(pod.Name, "testPredicate", predicateID, test.expectFit, test.expectedReasons, equivClass.hash, node)
}
fit, reasons, err := ecache.RunPredicate(test.pred.predicate, "testPredicate", pod, meta, node, equivClass, test.cache)
fit, reasons, err := nodeCache.RunPredicate(test.pred.predicate, "testPredicate", predicateID, pod, meta, node, equivClass)
if err != nil {
if err.Error() != test.expectedError {
@ -287,14 +266,14 @@ func TestRunPredicate(t *testing.T) {
if !test.expectCacheHit && test.pred.callCount == 0 {
t.Errorf("Predicate should be called")
}
_, _, invalid := ecache.lookupResult(pod.Name, node.Node().Name, "testPredicate", equivClass.hash)
if invalid && test.expectCacheWrite {
_, ok := nodeCache.lookupResult(pod.Name, node.Node().Name, "testPredicate", predicateID, equivClass.hash)
if !ok && test.expectCacheWrite {
t.Errorf("Cache write should happen")
}
if !test.expectCacheHit && test.expectCacheWrite && invalid {
if !test.expectCacheHit && test.expectCacheWrite && !ok {
t.Errorf("Cache write should happen")
}
if !test.expectCacheHit && !test.expectCacheWrite && !invalid {
if !test.expectCacheHit && !test.expectCacheWrite && ok {
t.Errorf("Cache write should not happen")
}
})
@ -302,80 +281,86 @@ func TestRunPredicate(t *testing.T) {
}
func TestUpdateResult(t *testing.T) {
predicatesOrdering := []string{"GeneralPredicates"}
tests := []struct {
name string
pod string
predicateKey string
predicateID int
nodeName string
fit bool
reasons []algorithm.PredicateFailureReason
equivalenceHash uint64
expectPredicateMap bool
expectCacheItem HostPredicate
cache schedulercache.Cache
expectCacheItem predicateResult
}{
{
name: "test 1",
pod: "testPod",
predicateKey: "GeneralPredicates",
predicateID: 0,
nodeName: "node1",
fit: true,
equivalenceHash: 123,
expectPredicateMap: false,
expectCacheItem: HostPredicate{
expectCacheItem: predicateResult{
Fit: true,
},
cache: &upToDateCache{},
},
{
name: "test 2",
pod: "testPod",
predicateKey: "GeneralPredicates",
predicateID: 0,
nodeName: "node2",
fit: false,
equivalenceHash: 123,
expectPredicateMap: true,
expectCacheItem: HostPredicate{
expectCacheItem: predicateResult{
Fit: false,
},
cache: &upToDateCache{},
},
}
for _, test := range tests {
ecache := NewEquivalenceCache()
if test.expectPredicateMap {
ecache.algorithmCache[test.nodeName] = AlgorithmCache{}
predicateItem := HostPredicate{
Fit: true,
}
ecache.algorithmCache[test.nodeName][test.predicateKey] =
PredicateMap{
test.equivalenceHash: predicateItem,
t.Run(test.name, func(t *testing.T) {
node := schedulercache.NewNodeInfo()
testNode := &v1.Node{ObjectMeta: metav1.ObjectMeta{Name: test.nodeName}}
node.SetNode(testNode)
// Initialize and populate equivalence class cache.
ecache := NewCache(predicatesOrdering)
nodeCache, _ := ecache.GetNodeCache(testNode.Name)
if test.expectPredicateMap {
predicateItem := predicateResult{
Fit: true,
}
}
node := schedulercache.NewNodeInfo()
node.SetNode(&v1.Node{ObjectMeta: metav1.ObjectMeta{Name: test.nodeName}})
ecache.updateResult(
test.pod,
test.predicateKey,
test.fit,
test.reasons,
test.equivalenceHash,
test.cache,
node,
)
cachedMapItem, ok := ecache.algorithmCache[test.nodeName][test.predicateKey]
if !ok {
t.Errorf("Failed: %s, can't find expected cache item: %v",
test.name, test.expectCacheItem)
} else {
if !reflect.DeepEqual(cachedMapItem[test.equivalenceHash], test.expectCacheItem) {
t.Errorf("Failed: %s, expected cached item: %v, but got: %v",
test.name, test.expectCacheItem, cachedMapItem[test.equivalenceHash])
nodeCache.cache[test.predicateID] =
resultMap{
test.equivalenceHash: predicateItem,
}
}
}
nodeCache.updateResult(
test.pod,
test.predicateKey,
test.predicateID,
test.fit,
test.reasons,
test.equivalenceHash,
node,
)
cachedMapItem := nodeCache.cache[test.predicateID]
if cachedMapItem == nil {
t.Errorf("can't find expected cache item: %v", test.expectCacheItem)
} else {
if !reflect.DeepEqual(cachedMapItem[test.equivalenceHash], test.expectCacheItem) {
t.Errorf("expected cached item: %v, but got: %v",
test.expectCacheItem, cachedMapItem[test.equivalenceHash])
}
}
})
}
}
@ -388,146 +373,157 @@ func slicesEqual(a, b []algorithm.PredicateFailureReason) bool {
}
func TestLookupResult(t *testing.T) {
predicatesOrdering := []string{"GeneralPredicates"}
tests := []struct {
name string
podName string
nodeName string
predicateKey string
predicateID int
equivalenceHashForUpdatePredicate uint64
equivalenceHashForCalPredicate uint64
cachedItem predicateItemType
expectedInvalidPredicateKey bool
expectedInvalidEquivalenceHash bool
expectedPredicateKeyMiss bool
expectedEquivalenceHashMiss bool
expectedPredicateItem predicateItemType
cache schedulercache.Cache
}{
{
name: "test 1",
podName: "testPod",
nodeName: "node1",
name: "test 1",
podName: "testPod",
nodeName: "node1",
equivalenceHashForUpdatePredicate: 123,
equivalenceHashForCalPredicate: 123,
predicateKey: "GeneralPredicates",
predicateID: 0,
cachedItem: predicateItemType{
fit: false,
reasons: []algorithm.PredicateFailureReason{predicates.ErrPodNotFitsHostPorts},
},
expectedInvalidPredicateKey: true,
expectedPredicateKeyMiss: true,
expectedPredicateItem: predicateItemType{
fit: false,
reasons: []algorithm.PredicateFailureReason{},
},
cache: &upToDateCache{},
},
{
name: "test 2",
podName: "testPod",
nodeName: "node2",
name: "test 2",
podName: "testPod",
nodeName: "node2",
equivalenceHashForUpdatePredicate: 123,
equivalenceHashForCalPredicate: 123,
predicateKey: "GeneralPredicates",
predicateID: 0,
cachedItem: predicateItemType{
fit: true,
},
expectedInvalidPredicateKey: false,
expectedPredicateKeyMiss: false,
expectedPredicateItem: predicateItemType{
fit: true,
reasons: []algorithm.PredicateFailureReason{},
},
cache: &upToDateCache{},
},
{
name: "test 3",
podName: "testPod",
nodeName: "node3",
name: "test 3",
podName: "testPod",
nodeName: "node3",
equivalenceHashForUpdatePredicate: 123,
equivalenceHashForCalPredicate: 123,
predicateKey: "GeneralPredicates",
predicateID: 0,
cachedItem: predicateItemType{
fit: false,
reasons: []algorithm.PredicateFailureReason{predicates.ErrPodNotFitsHostPorts},
},
expectedInvalidPredicateKey: false,
expectedPredicateKeyMiss: false,
expectedPredicateItem: predicateItemType{
fit: false,
reasons: []algorithm.PredicateFailureReason{predicates.ErrPodNotFitsHostPorts},
},
cache: &upToDateCache{},
},
{
name: "test 4",
podName: "testPod",
nodeName: "node4",
name: "test 4",
podName: "testPod",
nodeName: "node4",
equivalenceHashForUpdatePredicate: 123,
equivalenceHashForCalPredicate: 456,
predicateKey: "GeneralPredicates",
predicateID: 0,
cachedItem: predicateItemType{
fit: false,
reasons: []algorithm.PredicateFailureReason{predicates.ErrPodNotFitsHostPorts},
},
expectedInvalidPredicateKey: false,
expectedInvalidEquivalenceHash: true,
expectedPredicateKeyMiss: false,
expectedEquivalenceHashMiss: true,
expectedPredicateItem: predicateItemType{
fit: false,
reasons: []algorithm.PredicateFailureReason{},
},
cache: &upToDateCache{},
},
}
for _, test := range tests {
ecache := NewEquivalenceCache()
node := schedulercache.NewNodeInfo()
node.SetNode(&v1.Node{ObjectMeta: metav1.ObjectMeta{Name: test.nodeName}})
// set cached item to equivalence cache
ecache.updateResult(
test.podName,
test.predicateKey,
test.cachedItem.fit,
test.cachedItem.reasons,
test.equivalenceHashForUpdatePredicate,
test.cache,
node,
)
// if we want to do invalid, invalid the cached item
if test.expectedInvalidPredicateKey {
predicateKeys := sets.NewString()
predicateKeys.Insert(test.predicateKey)
ecache.InvalidateCachedPredicateItem(test.nodeName, predicateKeys)
}
// calculate predicate with equivalence cache
fit, reasons, invalid := ecache.lookupResult(test.podName,
test.nodeName,
test.predicateKey,
test.equivalenceHashForCalPredicate,
)
// returned invalid should match expectedInvalidPredicateKey or expectedInvalidEquivalenceHash
if test.equivalenceHashForUpdatePredicate != test.equivalenceHashForCalPredicate {
if invalid != test.expectedInvalidEquivalenceHash {
t.Errorf("Failed: %s, expected invalid: %v, but got: %v",
test.name, test.expectedInvalidEquivalenceHash, invalid)
t.Run(test.name, func(t *testing.T) {
testNode := &v1.Node{ObjectMeta: metav1.ObjectMeta{Name: test.nodeName}}
// Initialize and populate equivalence class cache.
ecache := NewCache(predicatesOrdering)
nodeCache, _ := ecache.GetNodeCache(testNode.Name)
node := schedulercache.NewNodeInfo()
node.SetNode(testNode)
// set cached item to equivalence cache
nodeCache.updateResult(
test.podName,
test.predicateKey,
test.predicateID,
test.cachedItem.fit,
test.cachedItem.reasons,
test.equivalenceHashForUpdatePredicate,
node,
)
// if we want to do invalid, invalid the cached item
if test.expectedPredicateKeyMiss {
predicateKeys := sets.NewString()
predicateKeys.Insert(test.predicateKey)
ecache.InvalidatePredicatesOnNode(test.nodeName, predicateKeys)
}
} else {
if invalid != test.expectedInvalidPredicateKey {
t.Errorf("Failed: %s, expected invalid: %v, but got: %v",
test.name, test.expectedInvalidPredicateKey, invalid)
// calculate predicate with equivalence cache
result, ok := nodeCache.lookupResult(test.podName,
test.nodeName,
test.predicateKey,
test.predicateID,
test.equivalenceHashForCalPredicate,
)
fit, reasons := result.Fit, result.FailReasons
// returned invalid should match expectedPredicateKeyMiss or expectedEquivalenceHashMiss
if test.equivalenceHashForUpdatePredicate != test.equivalenceHashForCalPredicate {
if ok && test.expectedEquivalenceHashMiss {
t.Errorf("Failed: %s, expected (equivalence hash) cache miss", test.name)
}
if !ok && !test.expectedEquivalenceHashMiss {
t.Errorf("Failed: %s, expected (equivalence hash) cache hit", test.name)
}
} else {
if ok && test.expectedPredicateKeyMiss {
t.Errorf("Failed: %s, expected (predicate key) cache miss", test.name)
}
if !ok && !test.expectedPredicateKeyMiss {
t.Errorf("Failed: %s, expected (predicate key) cache hit", test.name)
}
}
}
// returned predicate result should match expected predicate item
if fit != test.expectedPredicateItem.fit {
t.Errorf("Failed: %s, expected fit: %v, but got: %v", test.name, test.cachedItem.fit, fit)
}
if !slicesEqual(reasons, test.expectedPredicateItem.reasons) {
t.Errorf("Failed: %s, expected reasons: %v, but got: %v",
test.name, test.expectedPredicateItem.reasons, reasons)
}
// returned predicate result should match expected predicate item
if fit != test.expectedPredicateItem.fit {
t.Errorf("Failed: %s, expected fit: %v, but got: %v", test.name, test.cachedItem.fit, fit)
}
if !slicesEqual(reasons, test.expectedPredicateItem.reasons) {
t.Errorf("Failed: %s, expected reasons: %v, but got: %v",
test.name, test.expectedPredicateItem.reasons, reasons)
}
})
}
}
func TestGetEquivalenceHash(t *testing.T) {
ecache := NewEquivalenceCache()
pod1 := makeBasicPod("pod1")
pod2 := makeBasicPod("pod2")
@ -618,7 +614,7 @@ func TestGetEquivalenceHash(t *testing.T) {
t.Run(test.name, func(t *testing.T) {
for i, podInfo := range test.podInfoList {
testPod := podInfo.pod
eclassInfo := ecache.getEquivalenceClassInfo(testPod)
eclassInfo := NewClass(testPod)
if eclassInfo == nil && podInfo.hashIsValid {
t.Errorf("Failed: pod %v is expected to have valid hash", testPod)
}
@ -644,17 +640,20 @@ func TestGetEquivalenceHash(t *testing.T) {
func TestInvalidateCachedPredicateItemOfAllNodes(t *testing.T) {
testPredicate := "GeneralPredicates"
testPredicateID := 0
predicatesOrdering := []string{testPredicate}
// tests is used to initialize all nodes
tests := []struct {
name string
podName string
nodeName string
equivalenceHashForUpdatePredicate uint64
cachedItem predicateItemType
cache schedulercache.Cache
}{
{
podName: "testPod",
nodeName: "node1",
name: "hash predicate 123 not fits host ports",
podName: "testPod",
nodeName: "node1",
equivalenceHashForUpdatePredicate: 123,
cachedItem: predicateItemType{
fit: false,
@ -662,11 +661,11 @@ func TestInvalidateCachedPredicateItemOfAllNodes(t *testing.T) {
predicates.ErrPodNotFitsHostPorts,
},
},
cache: &upToDateCache{},
},
{
podName: "testPod",
nodeName: "node2",
name: "hash predicate 456 not fits host ports",
podName: "testPod",
nodeName: "node2",
equivalenceHashForUpdatePredicate: 456,
cachedItem: predicateItemType{
fit: false,
@ -674,217 +673,131 @@ func TestInvalidateCachedPredicateItemOfAllNodes(t *testing.T) {
predicates.ErrPodNotFitsHostPorts,
},
},
cache: &upToDateCache{},
},
{
podName: "testPod",
nodeName: "node3",
name: "hash predicate 123 fits",
podName: "testPod",
nodeName: "node3",
equivalenceHashForUpdatePredicate: 123,
cachedItem: predicateItemType{
fit: true,
},
cache: &upToDateCache{},
},
}
ecache := NewEquivalenceCache()
ecache := NewCache(predicatesOrdering)
for _, test := range tests {
node := schedulercache.NewNodeInfo()
node.SetNode(&v1.Node{ObjectMeta: metav1.ObjectMeta{Name: test.nodeName}})
testNode := &v1.Node{ObjectMeta: metav1.ObjectMeta{Name: test.nodeName}}
node.SetNode(testNode)
nodeCache, _ := ecache.GetNodeCache(testNode.Name)
// set cached item to equivalence cache
ecache.updateResult(
nodeCache.updateResult(
test.podName,
testPredicate,
testPredicateID,
test.cachedItem.fit,
test.cachedItem.reasons,
test.equivalenceHashForUpdatePredicate,
test.cache,
node,
)
}
// invalidate cached predicate for all nodes
ecache.InvalidateCachedPredicateItemOfAllNodes(sets.NewString(testPredicate))
ecache.InvalidatePredicates(sets.NewString(testPredicate))
// there should be no cached predicate any more
for _, test := range tests {
if algorithmCache, exist := ecache.algorithmCache[test.nodeName]; exist {
if _, exist := algorithmCache[testPredicate]; exist {
t.Errorf("Failed: cached item for predicate key: %v on node: %v should be invalidated",
testPredicate, test.nodeName)
break
t.Run(test.name, func(t *testing.T) {
if nodeCache, exist := ecache.nodeToCache[test.nodeName]; exist {
if cache := nodeCache.cache[testPredicateID]; cache != nil {
t.Errorf("Failed: cached item for predicate key: %v on node: %v should be invalidated",
testPredicate, test.nodeName)
}
}
}
})
}
}
func TestInvalidateAllCachedPredicateItemOfNode(t *testing.T) {
testPredicate := "GeneralPredicates"
testPredicateID := 0
predicatesOrdering := []string{testPredicate}
// tests is used to initialize all nodes
tests := []struct {
name string
podName string
nodeName string
equivalenceHashForUpdatePredicate uint64
cachedItem predicateItemType
cache schedulercache.Cache
}{
{
podName: "testPod",
nodeName: "node1",
name: "hash predicate 123 not fits host ports",
podName: "testPod",
nodeName: "node1",
equivalenceHashForUpdatePredicate: 123,
cachedItem: predicateItemType{
fit: false,
reasons: []algorithm.PredicateFailureReason{predicates.ErrPodNotFitsHostPorts},
},
cache: &upToDateCache{},
},
{
podName: "testPod",
nodeName: "node2",
name: "hash predicate 456 not fits host ports",
podName: "testPod",
nodeName: "node2",
equivalenceHashForUpdatePredicate: 456,
cachedItem: predicateItemType{
fit: false,
reasons: []algorithm.PredicateFailureReason{predicates.ErrPodNotFitsHostPorts},
},
cache: &upToDateCache{},
},
{
podName: "testPod",
nodeName: "node3",
name: "hash predicate 123 fits host ports",
podName: "testPod",
nodeName: "node3",
equivalenceHashForUpdatePredicate: 123,
cachedItem: predicateItemType{
fit: true,
},
cache: &upToDateCache{},
},
}
ecache := NewEquivalenceCache()
ecache := NewCache(predicatesOrdering)
for _, test := range tests {
node := schedulercache.NewNodeInfo()
node.SetNode(&v1.Node{ObjectMeta: metav1.ObjectMeta{Name: test.nodeName}})
testNode := &v1.Node{ObjectMeta: metav1.ObjectMeta{Name: test.nodeName}}
node.SetNode(testNode)
nodeCache, _ := ecache.GetNodeCache(testNode.Name)
// set cached item to equivalence cache
ecache.updateResult(
nodeCache.updateResult(
test.podName,
testPredicate,
testPredicateID,
test.cachedItem.fit,
test.cachedItem.reasons,
test.equivalenceHashForUpdatePredicate,
test.cache,
node,
)
}
for _, test := range tests {
// invalidate cached predicate for all nodes
ecache.InvalidateAllCachedPredicateItemOfNode(test.nodeName)
if _, exist := ecache.algorithmCache[test.nodeName]; exist {
t.Errorf("Failed: cached item for node: %v should be invalidated", test.nodeName)
break
}
t.Run(test.name, func(t *testing.T) {
oldNodeCache, _ := ecache.GetNodeCache(test.nodeName)
oldGeneration := oldNodeCache.generation
// invalidate cached predicate for all nodes
ecache.InvalidateAllPredicatesOnNode(test.nodeName)
if n, _ := ecache.GetNodeCache(test.nodeName); oldGeneration == n.generation {
t.Errorf("Failed: cached item for node: %v should be invalidated", test.nodeName)
}
})
}
}
func BenchmarkEquivalenceHash(b *testing.B) {
pod := makeBasicPod("test")
for i := 0; i < b.N; i++ {
getEquivalenceHash(pod)
}
}
// syncingMockCache delegates method calls to an actual Cache,
// but calls to UpdateNodeNameToInfoMap synchronize with the test.
type syncingMockCache struct {
schedulercache.Cache
cycleStart, cacheInvalidated chan struct{}
once sync.Once
}
// UpdateNodeNameToInfoMap delegates to the real implementation, but on the first call, it
// synchronizes with the test.
//
// Since UpdateNodeNameToInfoMap is one of the first steps of (*genericScheduler).Schedule, we use
// this point to signal to the test that a scheduling cycle has started.
func (c *syncingMockCache) UpdateNodeNameToInfoMap(infoMap map[string]*schedulercache.NodeInfo) error {
err := c.Cache.UpdateNodeNameToInfoMap(infoMap)
c.once.Do(func() {
c.cycleStart <- struct{}{}
<-c.cacheInvalidated
})
return err
}
// TestEquivalenceCacheInvalidationRace tests that equivalence cache invalidation is correctly
// handled when an invalidation event happens early in a scheduling cycle. Specifically, the event
// occurs after schedulercache is snapshotted and before equivalence cache lock is acquired.
func TestEquivalenceCacheInvalidationRace(t *testing.T) {
// Create a predicate that returns false the first time and true on subsequent calls.
podWillFit := false
var callCount int
testPredicate := func(pod *v1.Pod,
meta algorithm.PredicateMetadata,
nodeInfo *schedulercache.NodeInfo) (bool, []algorithm.PredicateFailureReason, error) {
callCount++
if !podWillFit {
podWillFit = true
return false, []algorithm.PredicateFailureReason{predicates.ErrFakePredicate}, nil
}
return true, nil, nil
}
// Set up the mock cache.
cache := schedulercache.New(time.Duration(0), wait.NeverStop)
cache.AddNode(&v1.Node{ObjectMeta: metav1.ObjectMeta{Name: "machine1"}})
mockCache := &syncingMockCache{
Cache: cache,
cycleStart: make(chan struct{}),
cacheInvalidated: make(chan struct{}),
}
eCache := NewEquivalenceCache()
// Ensure that equivalence cache invalidation happens after the scheduling cycle starts, but before
// the equivalence cache would be updated.
go func() {
<-mockCache.cycleStart
pod := &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: "new-pod", UID: "new-pod"},
Spec: v1.PodSpec{NodeName: "machine1"}}
if err := cache.AddPod(pod); err != nil {
t.Errorf("Could not add pod to cache: %v", err)
}
eCache.InvalidateAllCachedPredicateItemOfNode("machine1")
mockCache.cacheInvalidated <- struct{}{}
}()
// Set up the scheduler.
ps := map[string]algorithm.FitPredicate{"testPredicate": testPredicate}
predicates.SetPredicatesOrdering([]string{"testPredicate"})
prioritizers := []algorithm.PriorityConfig{{Map: EqualPriorityMap, Weight: 1}}
pvcLister := schedulertesting.FakePersistentVolumeClaimLister([]*v1.PersistentVolumeClaim{})
scheduler := NewGenericScheduler(
mockCache,
eCache,
NewSchedulingQueue(),
ps,
algorithm.EmptyPredicateMetadataProducer,
prioritizers,
algorithm.EmptyPriorityMetadataProducer,
nil, nil, pvcLister, true, false)
// First scheduling attempt should fail.
nodeLister := schedulertesting.FakeNodeLister(makeNodeList([]string{"machine1"}))
pod := &v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "test-pod"}}
machine, err := scheduler.Schedule(pod, nodeLister)
if machine != "" || err == nil {
t.Error("First scheduling attempt did not fail")
}
// Second scheduling attempt should succeed because cache was invalidated.
_, err = scheduler.Schedule(pod, nodeLister)
if err != nil {
t.Errorf("Second scheduling attempt failed: %v", err)
}
if callCount != 2 {
t.Errorf("Predicate should have been called twice. Was called %d times.", callCount)
getEquivalencePod(pod)
}
}

View File

@ -1,307 +0,0 @@
/*
Copyright 2016 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 core
import (
"fmt"
"hash/fnv"
"sync"
"k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/kubernetes/pkg/scheduler/algorithm"
"k8s.io/kubernetes/pkg/scheduler/algorithm/predicates"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
hashutil "k8s.io/kubernetes/pkg/util/hash"
"github.com/golang/glog"
)
// EquivalenceCache holds:
// 1. a map of AlgorithmCache with node name as key
// 2. function to get equivalence pod
type EquivalenceCache struct {
mu sync.RWMutex
algorithmCache map[string]AlgorithmCache
}
// The AlgorithmCache stores PredicateMap with predicate name as key, PredicateMap as value.
type AlgorithmCache map[string]PredicateMap
// PredicateMap stores HostPrediacte with equivalence hash as key
type PredicateMap map[uint64]HostPredicate
// HostPredicate is the cached predicate result
type HostPredicate struct {
Fit bool
FailReasons []algorithm.PredicateFailureReason
}
// NewEquivalenceCache returns EquivalenceCache to speed up predicates by caching
// result from previous scheduling.
func NewEquivalenceCache() *EquivalenceCache {
return &EquivalenceCache{
algorithmCache: make(map[string]AlgorithmCache),
}
}
// RunPredicate will return a cached predicate result. In case of a cache miss, the predicate will
// be run and its results cached for the next call.
//
// NOTE: RunPredicate will not update the equivalence cache if the given NodeInfo is stale.
func (ec *EquivalenceCache) RunPredicate(
pred algorithm.FitPredicate,
predicateKey string,
pod *v1.Pod,
meta algorithm.PredicateMetadata,
nodeInfo *schedulercache.NodeInfo,
equivClassInfo *equivalenceClassInfo,
cache schedulercache.Cache,
) (bool, []algorithm.PredicateFailureReason, error) {
if nodeInfo == nil || nodeInfo.Node() == nil {
// This may happen during tests.
return false, []algorithm.PredicateFailureReason{}, fmt.Errorf("nodeInfo is nil or node is invalid")
}
fit, reasons, invalid := ec.lookupResult(pod.GetName(), nodeInfo.Node().GetName(), predicateKey, equivClassInfo.hash)
if !invalid {
return fit, reasons, nil
}
fit, reasons, err := pred(pod, meta, nodeInfo)
if err != nil {
return fit, reasons, err
}
if cache != nil {
ec.updateResult(pod.GetName(), predicateKey, fit, reasons, equivClassInfo.hash, cache, nodeInfo)
}
return fit, reasons, nil
}
// updateResult updates the cached result of a predicate.
func (ec *EquivalenceCache) updateResult(
podName, predicateKey string,
fit bool,
reasons []algorithm.PredicateFailureReason,
equivalenceHash uint64,
cache schedulercache.Cache,
nodeInfo *schedulercache.NodeInfo,
) {
ec.mu.Lock()
defer ec.mu.Unlock()
if nodeInfo == nil || nodeInfo.Node() == nil {
// This may happen during tests.
return
}
// Skip update if NodeInfo is stale.
if !cache.IsUpToDate(nodeInfo) {
return
}
nodeName := nodeInfo.Node().GetName()
if _, exist := ec.algorithmCache[nodeName]; !exist {
ec.algorithmCache[nodeName] = AlgorithmCache{}
}
predicateItem := HostPredicate{
Fit: fit,
FailReasons: reasons,
}
// if cached predicate map already exists, just update the predicate by key
if predicateMap, ok := ec.algorithmCache[nodeName][predicateKey]; ok {
// maps in golang are references, no need to add them back
predicateMap[equivalenceHash] = predicateItem
} else {
ec.algorithmCache[nodeName][predicateKey] =
PredicateMap{
equivalenceHash: predicateItem,
}
}
glog.V(5).Infof("Updated cached predicate: %v for pod: %v on node: %s, with item %v", predicateKey, podName, nodeName, predicateItem)
}
// lookupResult returns cached predicate results:
// 1. if pod fit
// 2. reasons if pod did not fit
// 3. if cache item is not found
func (ec *EquivalenceCache) lookupResult(
podName, nodeName, predicateKey string,
equivalenceHash uint64,
) (bool, []algorithm.PredicateFailureReason, bool) {
ec.mu.RLock()
defer ec.mu.RUnlock()
glog.V(5).Infof("Begin to calculate predicate: %v for pod: %s on node: %s based on equivalence cache",
predicateKey, podName, nodeName)
if hostPredicate, exist := ec.algorithmCache[nodeName][predicateKey][equivalenceHash]; exist {
if hostPredicate.Fit {
return true, []algorithm.PredicateFailureReason{}, false
}
return false, hostPredicate.FailReasons, false
}
// is invalid
return false, []algorithm.PredicateFailureReason{}, true
}
// InvalidateCachedPredicateItem marks all items of given predicateKeys, of all pods, on the given node as invalid
func (ec *EquivalenceCache) InvalidateCachedPredicateItem(nodeName string, predicateKeys sets.String) {
if len(predicateKeys) == 0 {
return
}
ec.mu.Lock()
defer ec.mu.Unlock()
for predicateKey := range predicateKeys {
delete(ec.algorithmCache[nodeName], predicateKey)
}
glog.V(5).Infof("Done invalidating cached predicates: %v on node: %s", predicateKeys, nodeName)
}
// InvalidateCachedPredicateItemOfAllNodes marks all items of given predicateKeys, of all pods, on all node as invalid
func (ec *EquivalenceCache) InvalidateCachedPredicateItemOfAllNodes(predicateKeys sets.String) {
if len(predicateKeys) == 0 {
return
}
ec.mu.Lock()
defer ec.mu.Unlock()
// algorithmCache uses nodeName as key, so we just iterate it and invalid given predicates
for _, algorithmCache := range ec.algorithmCache {
for predicateKey := range predicateKeys {
delete(algorithmCache, predicateKey)
}
}
glog.V(5).Infof("Done invalidating cached predicates: %v on all node", predicateKeys)
}
// InvalidateAllCachedPredicateItemOfNode marks all cached items on given node as invalid
func (ec *EquivalenceCache) InvalidateAllCachedPredicateItemOfNode(nodeName string) {
ec.mu.Lock()
defer ec.mu.Unlock()
delete(ec.algorithmCache, nodeName)
glog.V(5).Infof("Done invalidating all cached predicates on node: %s", nodeName)
}
// InvalidateCachedPredicateItemForPodAdd is a wrapper of InvalidateCachedPredicateItem for pod add case
func (ec *EquivalenceCache) InvalidateCachedPredicateItemForPodAdd(pod *v1.Pod, nodeName string) {
// MatchInterPodAffinity: we assume scheduler can make sure newly bound pod
// will not break the existing inter pod affinity. So we does not need to invalidate
// MatchInterPodAffinity when pod added.
//
// But when a pod is deleted, existing inter pod affinity may become invalid.
// (e.g. this pod was preferred by some else, or vice versa)
//
// NOTE: assumptions above will not stand when we implemented features like
// RequiredDuringSchedulingRequiredDuringExecution.
// NoDiskConflict: the newly scheduled pod fits to existing pods on this node,
// it will also fits to equivalence class of existing pods
// GeneralPredicates: will always be affected by adding a new pod
invalidPredicates := sets.NewString(predicates.GeneralPred)
// MaxPDVolumeCountPredicate: we check the volumes of pod to make decision.
for _, vol := range pod.Spec.Volumes {
if vol.PersistentVolumeClaim != nil {
invalidPredicates.Insert(predicates.MaxEBSVolumeCountPred, predicates.MaxGCEPDVolumeCountPred, predicates.MaxAzureDiskVolumeCountPred)
} else {
if vol.AWSElasticBlockStore != nil {
invalidPredicates.Insert(predicates.MaxEBSVolumeCountPred)
}
if vol.GCEPersistentDisk != nil {
invalidPredicates.Insert(predicates.MaxGCEPDVolumeCountPred)
}
if vol.AzureDisk != nil {
invalidPredicates.Insert(predicates.MaxAzureDiskVolumeCountPred)
}
}
}
ec.InvalidateCachedPredicateItem(nodeName, invalidPredicates)
}
// equivalenceClassInfo holds equivalence hash which is used for checking equivalence cache.
// We will pass this to podFitsOnNode to ensure equivalence hash is only calculated per schedule.
type equivalenceClassInfo struct {
// Equivalence hash.
hash uint64
}
// getEquivalenceClassInfo returns a hash of the given pod.
// The hashing function returns the same value for any two pods that are
// equivalent from the perspective of scheduling.
func (ec *EquivalenceCache) getEquivalenceClassInfo(pod *v1.Pod) *equivalenceClassInfo {
equivalencePod := getEquivalenceHash(pod)
if equivalencePod != nil {
hash := fnv.New32a()
hashutil.DeepHashObject(hash, equivalencePod)
return &equivalenceClassInfo{
hash: uint64(hash.Sum32()),
}
}
return nil
}
// equivalencePod is the set of pod attributes which must match for two pods to
// be considered equivalent for scheduling purposes. For correctness, this must
// include any Pod field which is used by a FitPredicate.
//
// NOTE: For equivalence hash to be formally correct, lists and maps in the
// equivalencePod should be normalized. (e.g. by sorting them) However, the
// vast majority of equivalent pod classes are expected to be created from a
// single pod template, so they will all have the same ordering.
type equivalencePod struct {
Namespace *string
Labels map[string]string
Affinity *v1.Affinity
Containers []v1.Container // See note about ordering
InitContainers []v1.Container // See note about ordering
NodeName *string
NodeSelector map[string]string
Tolerations []v1.Toleration
Volumes []v1.Volume // See note about ordering
}
// getEquivalenceHash returns the equivalencePod for a Pod.
func getEquivalenceHash(pod *v1.Pod) *equivalencePod {
ep := &equivalencePod{
Namespace: &pod.Namespace,
Labels: pod.Labels,
Affinity: pod.Spec.Affinity,
Containers: pod.Spec.Containers,
InitContainers: pod.Spec.InitContainers,
NodeName: &pod.Spec.NodeName,
NodeSelector: pod.Spec.NodeSelector,
Tolerations: pod.Spec.Tolerations,
Volumes: pod.Spec.Volumes,
}
// DeepHashObject considers nil and empty slices to be different. Normalize them.
if len(ep.Containers) == 0 {
ep.Containers = nil
}
if len(ep.InitContainers) == 0 {
ep.InitContainers = nil
}
if len(ep.Tolerations) == 0 {
ep.Tolerations = nil
}
if len(ep.Volumes) == 0 {
ep.Volumes = nil
}
// Normalize empty maps also.
if len(ep.Labels) == 0 {
ep.Labels = nil
}
if len(ep.NodeSelector) == 0 {
ep.NodeSelector = nil
}
// TODO(misterikkit): Also normalize nested maps and slices.
return ep
}

View File

@ -107,6 +107,11 @@ func NewHTTPExtender(config *schedulerapi.ExtenderConfig) (algorithm.SchedulerEx
}, nil
}
// Name returns extenderURL to identifies the extender.
func (h *HTTPExtender) Name() string {
return h.extenderURL
}
// IsIgnorable returns true indicates scheduling should not fail when this extender
// is unavailable
func (h *HTTPExtender) IsIgnorable() bool {
@ -138,7 +143,7 @@ func (h *HTTPExtender) ProcessPreemption(
// If extender has cached node info, pass NodeNameToMetaVictims in args.
nodeNameToMetaVictims := convertToNodeNameToMetaVictims(nodeToVictims)
args = &schedulerapi.ExtenderPreemptionArgs{
Pod: pod,
Pod: pod,
NodeNameToMetaVictims: nodeNameToMetaVictims,
}
} else {

View File

@ -27,6 +27,8 @@ import (
"k8s.io/kubernetes/pkg/scheduler/algorithm"
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
schedulerinternalcache "k8s.io/kubernetes/pkg/scheduler/internal/cache"
internalqueue "k8s.io/kubernetes/pkg/scheduler/internal/queue"
schedulertesting "k8s.io/kubernetes/pkg/scheduler/testing"
"k8s.io/kubernetes/pkg/scheduler/util"
)
@ -118,6 +120,10 @@ type FakeExtender struct {
cachedNodeNameToInfo map[string]*schedulercache.NodeInfo
}
func (f *FakeExtender) Name() string {
return "FakeExtender"
}
func (f *FakeExtender) IsIgnorable() bool {
return f.ignorable
}
@ -454,7 +460,7 @@ func TestGenericSchedulerWithExtenders(t *testing.T) {
// because of the errors from errorPredicateExtender and/or
// errorPrioritizerExtender.
predicates: map[string]algorithm.FitPredicate{"true": truePredicate},
prioritizers: []algorithm.PriorityConfig{{Map: EqualPriorityMap, Weight: 1}},
prioritizers: []algorithm.PriorityConfig{{Function: machine2Prioritizer, Weight: 1}},
extenders: []FakeExtender{
{
predicates: []fitPredicate{errorPredicateExtender},
@ -492,43 +498,47 @@ func TestGenericSchedulerWithExtenders(t *testing.T) {
}
for _, test := range tests {
extenders := []algorithm.SchedulerExtender{}
for ii := range test.extenders {
extenders = append(extenders, &test.extenders[ii])
}
cache := schedulercache.New(time.Duration(0), wait.NeverStop)
for _, name := range test.nodes {
cache.AddNode(createNode(name))
}
queue := NewSchedulingQueue()
scheduler := NewGenericScheduler(
cache,
nil,
queue,
test.predicates,
algorithm.EmptyPredicateMetadataProducer,
test.prioritizers,
algorithm.EmptyPriorityMetadataProducer,
extenders,
nil,
schedulertesting.FakePersistentVolumeClaimLister{},
false,
false)
podIgnored := &v1.Pod{}
machine, err := scheduler.Schedule(podIgnored, schedulertesting.FakeNodeLister(makeNodeList(test.nodes)))
if test.expectsErr {
if err == nil {
t.Errorf("Unexpected non-error for %s, machine %s", test.name, machine)
t.Run(test.name, func(t *testing.T) {
extenders := []algorithm.SchedulerExtender{}
for ii := range test.extenders {
extenders = append(extenders, &test.extenders[ii])
}
} else {
if err != nil {
t.Errorf("Unexpected error: %v", err)
continue
cache := schedulerinternalcache.New(time.Duration(0), wait.NeverStop)
for _, name := range test.nodes {
cache.AddNode(createNode(name))
}
if test.expectedHost != machine {
t.Errorf("Failed : %s, Expected: %s, Saw: %s", test.name, test.expectedHost, machine)
queue := internalqueue.NewSchedulingQueue()
scheduler := NewGenericScheduler(
cache,
nil,
queue,
test.predicates,
algorithm.EmptyPredicateMetadataProducer,
test.prioritizers,
algorithm.EmptyPriorityMetadataProducer,
extenders,
nil,
schedulertesting.FakePersistentVolumeClaimLister{},
schedulertesting.FakePDBLister{},
false,
false,
schedulerapi.DefaultPercentageOfNodesToScore)
podIgnored := &v1.Pod{}
machine, err := scheduler.Schedule(podIgnored, schedulertesting.FakeNodeLister(makeNodeList(test.nodes)))
if test.expectsErr {
if err == nil {
t.Errorf("Unexpected non-error, machine %s", machine)
}
} else {
if err != nil {
t.Errorf("Unexpected error: %v", err)
return
}
if test.expectedHost != machine {
t.Errorf("Expected: %s, Saw: %s", test.expectedHost, machine)
}
}
}
})
}
}

View File

@ -17,6 +17,7 @@ limitations under the License.
package core
import (
"context"
"fmt"
"math"
"sort"
@ -25,7 +26,7 @@ import (
"sync/atomic"
"time"
"github.com/golang/glog"
"k8s.io/klog"
"k8s.io/api/core/v1"
policy "k8s.io/api/policy/v1beta1"
@ -39,11 +40,22 @@ import (
"k8s.io/kubernetes/pkg/scheduler/algorithm/predicates"
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
"k8s.io/kubernetes/pkg/scheduler/core/equivalence"
schedulerinternalcache "k8s.io/kubernetes/pkg/scheduler/internal/cache"
internalqueue "k8s.io/kubernetes/pkg/scheduler/internal/queue"
"k8s.io/kubernetes/pkg/scheduler/metrics"
"k8s.io/kubernetes/pkg/scheduler/util"
"k8s.io/kubernetes/pkg/scheduler/volumebinder"
)
const (
// minFeasibleNodesToFind is the minimum number of nodes that would be scored
// in each scheduling cycle. This is a semi-arbitrary value to ensure that a
// certain minimum of nodes are checked for feasibility. This in turn helps
// ensure a minimum level of spreading.
minFeasibleNodesToFind = 100
)
// FailedPredicateMap declares a map[string][]algorithm.PredicateFailureReason type.
type FailedPredicateMap map[string][]algorithm.PredicateFailureReason
@ -84,21 +96,41 @@ func (f *FitError) Error() string {
}
type genericScheduler struct {
cache schedulercache.Cache
equivalenceCache *EquivalenceCache
schedulingQueue SchedulingQueue
cache schedulerinternalcache.Cache
equivalenceCache *equivalence.Cache
schedulingQueue internalqueue.SchedulingQueue
predicates map[string]algorithm.FitPredicate
priorityMetaProducer algorithm.PriorityMetadataProducer
predicateMetaProducer algorithm.PredicateMetadataProducer
prioritizers []algorithm.PriorityConfig
extenders []algorithm.SchedulerExtender
lastNodeIndexLock sync.Mutex
lastNodeIndex uint64
alwaysCheckAllPredicates bool
cachedNodeInfoMap map[string]*schedulercache.NodeInfo
volumeBinder *volumebinder.VolumeBinder
pvcLister corelisters.PersistentVolumeClaimLister
pdbLister algorithm.PDBLister
disablePreemption bool
percentageOfNodesToScore int32
}
// snapshot snapshots equivalane cache and node infos for all fit and priority
// functions.
func (g *genericScheduler) snapshot() error {
// IMPORTANT NOTE: We must snapshot equivalence cache before snapshotting
// scheduler cache, otherwise stale data may be written into equivalence
// cache, e.g.
// 1. snapshot cache
// 2. event arrives, updating cache and invalidating predicates or whole node cache
// 3. snapshot ecache
// 4. evaludate predicates
// 5. stale result will be written to ecache
if g.equivalenceCache != nil {
g.equivalenceCache.Snapshot()
}
// Used for all fit and priority funcs.
return g.cache.UpdateNodeNameToInfoMap(g.cachedNodeInfoMap)
}
// Schedule tries to schedule the given pod to one of the nodes in the node list.
@ -120,8 +152,7 @@ func (g *genericScheduler) Schedule(pod *v1.Pod, nodeLister algorithm.NodeLister
return "", ErrNoNodesAvailable
}
// Used for all fit and priority funcs.
err = g.cache.UpdateNodeNameToInfoMap(g.cachedNodeInfoMap)
err = g.snapshot()
if err != nil {
return "", err
}
@ -175,6 +206,22 @@ func (g *genericScheduler) Predicates() map[string]algorithm.FitPredicate {
return g.predicates
}
// findMaxScores returns the indexes of nodes in the "priorityList" that has the highest "Score".
func findMaxScores(priorityList schedulerapi.HostPriorityList) []int {
maxScoreIndexes := make([]int, 0, len(priorityList)/2)
maxScore := priorityList[0].Score
for i, hp := range priorityList {
if hp.Score > maxScore {
maxScore = hp.Score
maxScoreIndexes = maxScoreIndexes[:0]
maxScoreIndexes = append(maxScoreIndexes, i)
} else if hp.Score == maxScore {
maxScoreIndexes = append(maxScoreIndexes, i)
}
}
return maxScoreIndexes
}
// selectHost takes a prioritized list of nodes and then picks one
// in a round-robin manner from the nodes that had the highest score.
func (g *genericScheduler) selectHost(priorityList schedulerapi.HostPriorityList) (string, error) {
@ -182,16 +229,11 @@ func (g *genericScheduler) selectHost(priorityList schedulerapi.HostPriorityList
return "", fmt.Errorf("empty priorityList")
}
sort.Sort(sort.Reverse(priorityList))
maxScore := priorityList[0].Score
firstAfterMaxScore := sort.Search(len(priorityList), func(i int) bool { return priorityList[i].Score < maxScore })
g.lastNodeIndexLock.Lock()
ix := int(g.lastNodeIndex % uint64(firstAfterMaxScore))
maxScores := findMaxScores(priorityList)
ix := int(g.lastNodeIndex % uint64(len(maxScores)))
g.lastNodeIndex++
g.lastNodeIndexLock.Unlock()
return priorityList[ix].Host, nil
return priorityList[maxScores[ix]].Host, nil
}
// preempt finds nodes with pods that can be preempted to make room for "pod" to
@ -206,12 +248,12 @@ func (g *genericScheduler) Preempt(pod *v1.Pod, nodeLister algorithm.NodeLister,
if !ok || fitError == nil {
return nil, nil, nil, nil
}
err := g.cache.UpdateNodeNameToInfoMap(g.cachedNodeInfoMap)
err := g.snapshot()
if err != nil {
return nil, nil, nil, err
}
if !podEligibleToPreemptOthers(pod, g.cachedNodeInfoMap) {
glog.V(5).Infof("Pod %v is not eligible for more preemption.", pod.Name)
klog.V(5).Infof("Pod %v/%v is not eligible for more preemption.", pod.Namespace, pod.Name)
return nil, nil, nil, nil
}
allNodes, err := nodeLister.List()
@ -221,13 +263,13 @@ func (g *genericScheduler) Preempt(pod *v1.Pod, nodeLister algorithm.NodeLister,
if len(allNodes) == 0 {
return nil, nil, nil, ErrNoNodesAvailable
}
potentialNodes := nodesWherePreemptionMightHelp(pod, allNodes, fitError.FailedPredicates)
potentialNodes := nodesWherePreemptionMightHelp(allNodes, fitError.FailedPredicates)
if len(potentialNodes) == 0 {
glog.V(3).Infof("Preemption will not help schedule pod %v on any node.", pod.Name)
klog.V(3).Infof("Preemption will not help schedule pod %v/%v on any node.", pod.Namespace, pod.Name)
// In this case, we should clean-up any existing nominated node name of the pod.
return nil, nil, []*v1.Pod{pod}, nil
}
pdbs, err := g.cache.ListPDBs(labels.Everything())
pdbs, err := g.pdbLister.List(labels.Everything())
if err != nil {
return nil, nil, nil, err
}
@ -271,7 +313,7 @@ func (g *genericScheduler) processPreemptionWithExtenders(
) (map[*v1.Node]*schedulerapi.Victims, error) {
if len(nodeToVictims) > 0 {
for _, extender := range g.extenders {
if extender.SupportsPreemption() {
if extender.SupportsPreemption() && extender.IsInterested(pod) {
newNodeToVictims, err := extender.ProcessPreemption(
pod,
nodeToVictims,
@ -279,7 +321,7 @@ func (g *genericScheduler) processPreemptionWithExtenders(
)
if err != nil {
if extender.IsIgnorable() {
glog.Warningf("Skipping extender %v as it returned error %v and has ignorable flag set",
klog.Warningf("Skipping extender %v as it returned error %v and has ignorable flag set",
extender, err)
continue
}
@ -325,6 +367,20 @@ func (g *genericScheduler) getLowerPriorityNominatedPods(pod *v1.Pod, nodeName s
return lowerPriorityPods
}
// numFeasibleNodesToFind returns the number of feasible nodes that once found, the scheduler stops
// its search for more feasible nodes.
func (g *genericScheduler) numFeasibleNodesToFind(numAllNodes int32) int32 {
if numAllNodes < minFeasibleNodesToFind || g.percentageOfNodesToScore <= 0 ||
g.percentageOfNodesToScore >= 100 {
return numAllNodes
}
numNodes := numAllNodes * g.percentageOfNodesToScore / 100
if numNodes < minFeasibleNodesToFind {
return minFeasibleNodesToFind
}
return numNodes
}
// Filters the nodes to find the ones that fit based on the given predicate functions
// Each node is passed through the predicate functions to determine if it is a fit
func (g *genericScheduler) findNodesThatFit(pod *v1.Pod, nodes []*v1.Node) ([]*v1.Node, FailedPredicateMap, error) {
@ -334,34 +390,44 @@ func (g *genericScheduler) findNodesThatFit(pod *v1.Pod, nodes []*v1.Node) ([]*v
if len(g.predicates) == 0 {
filtered = nodes
} else {
allNodes := int32(g.cache.NodeTree().NumNodes)
numNodesToFind := g.numFeasibleNodesToFind(allNodes)
// Create filtered list with enough space to avoid growing it
// and allow assigning.
filtered = make([]*v1.Node, len(nodes))
filtered = make([]*v1.Node, numNodesToFind)
errs := errors.MessageCountMap{}
var predicateResultLock sync.Mutex
var filteredLen int32
var (
predicateResultLock sync.Mutex
filteredLen int32
equivClass *equivalence.Class
)
ctx, cancel := context.WithCancel(context.Background())
// We can use the same metadata producer for all nodes.
meta := g.predicateMetaProducer(pod, g.cachedNodeInfoMap)
var equivCacheInfo *equivalenceClassInfo
if g.equivalenceCache != nil {
// getEquivalenceClassInfo will return immediately if no equivalence pod found
equivCacheInfo = g.equivalenceCache.getEquivalenceClassInfo(pod)
equivClass = equivalence.NewClass(pod)
}
checkNode := func(i int) {
nodeName := nodes[i].Name
var nodeCache *equivalence.NodeCache
nodeName := g.cache.NodeTree().Next()
if g.equivalenceCache != nil {
nodeCache = g.equivalenceCache.LoadNodeCache(nodeName)
}
fits, failedPredicates, err := podFitsOnNode(
pod,
meta,
g.cachedNodeInfoMap[nodeName],
g.predicates,
g.cache,
g.equivalenceCache,
nodeCache,
g.schedulingQueue,
g.alwaysCheckAllPredicates,
equivCacheInfo,
equivClass,
)
if err != nil {
predicateResultLock.Lock()
@ -370,14 +436,24 @@ func (g *genericScheduler) findNodesThatFit(pod *v1.Pod, nodes []*v1.Node) ([]*v
return
}
if fits {
filtered[atomic.AddInt32(&filteredLen, 1)-1] = nodes[i]
length := atomic.AddInt32(&filteredLen, 1)
if length > numNodesToFind {
cancel()
atomic.AddInt32(&filteredLen, -1)
} else {
filtered[length-1] = g.cachedNodeInfoMap[nodeName].Node()
}
} else {
predicateResultLock.Lock()
failedPredicateMap[nodeName] = failedPredicates
predicateResultLock.Unlock()
}
}
workqueue.Parallelize(16, len(nodes), checkNode)
// Stops searching for more nodes once the configured number of feasible nodes
// are found.
workqueue.ParallelizeUntil(ctx, 16, int(allNodes), checkNode)
filtered = filtered[:filteredLen]
if len(errs) > 0 {
return []*v1.Node{}, FailedPredicateMap{}, errors.CreateAggregateFromMessageCountMap(errs)
@ -392,7 +468,7 @@ func (g *genericScheduler) findNodesThatFit(pod *v1.Pod, nodes []*v1.Node) ([]*v
filteredList, failedMap, err := extender.Filter(pod, filtered, g.cachedNodeInfoMap)
if err != nil {
if extender.IsIgnorable() {
glog.Warningf("Skipping extender %v as it returned error %v and has ignorable flag set",
klog.Warningf("Skipping extender %v as it returned error %v and has ignorable flag set",
extender, err)
continue
} else {
@ -418,8 +494,8 @@ func (g *genericScheduler) findNodesThatFit(pod *v1.Pod, nodes []*v1.Node) ([]*v
// addNominatedPods adds pods with equal or greater priority which are nominated
// to run on the node given in nodeInfo to meta and nodeInfo. It returns 1) whether
// any pod was found, 2) augmented meta data, 3) augmented nodeInfo.
func addNominatedPods(podPriority int32, meta algorithm.PredicateMetadata,
nodeInfo *schedulercache.NodeInfo, queue SchedulingQueue) (bool, algorithm.PredicateMetadata,
func addNominatedPods(pod *v1.Pod, meta algorithm.PredicateMetadata,
nodeInfo *schedulercache.NodeInfo, queue internalqueue.SchedulingQueue) (bool, algorithm.PredicateMetadata,
*schedulercache.NodeInfo) {
if queue == nil || nodeInfo == nil || nodeInfo.Node() == nil {
// This may happen only in tests.
@ -435,7 +511,7 @@ func addNominatedPods(podPriority int32, meta algorithm.PredicateMetadata,
}
nodeInfoOut := nodeInfo.Clone()
for _, p := range nominatedPods {
if util.GetPodPriority(p) >= podPriority {
if util.GetPodPriority(p) >= util.GetPodPriority(pod) && p.UID != pod.UID {
nodeInfoOut.AddPod(p)
if metaOut != nil {
metaOut.AddPod(p, nodeInfoOut)
@ -460,11 +536,10 @@ func podFitsOnNode(
meta algorithm.PredicateMetadata,
info *schedulercache.NodeInfo,
predicateFuncs map[string]algorithm.FitPredicate,
cache schedulercache.Cache,
ecache *EquivalenceCache,
queue SchedulingQueue,
nodeCache *equivalence.NodeCache,
queue internalqueue.SchedulingQueue,
alwaysCheckAllPredicates bool,
equivCacheInfo *equivalenceClassInfo,
equivClass *equivalence.Class,
) (bool, []algorithm.PredicateFailureReason, error) {
var (
eCacheAvailable bool
@ -494,15 +569,15 @@ func podFitsOnNode(
metaToUse := meta
nodeInfoToUse := info
if i == 0 {
podsAdded, metaToUse, nodeInfoToUse = addNominatedPods(util.GetPodPriority(pod), meta, info, queue)
podsAdded, metaToUse, nodeInfoToUse = addNominatedPods(pod, meta, info, queue)
} else if !podsAdded || len(failedPredicates) != 0 {
break
}
// Bypass eCache if node has any nominated pods.
// TODO(bsalamat): consider using eCache and adding proper eCache invalidations
// when pods are nominated or their nominations change.
eCacheAvailable = equivCacheInfo != nil && !podsAdded
for _, predicateKey := range predicates.Ordering() {
eCacheAvailable = equivClass != nil && nodeCache != nil && !podsAdded
for predicateID, predicateKey := range predicates.Ordering() {
var (
fit bool
reasons []algorithm.PredicateFailureReason
@ -511,7 +586,7 @@ func podFitsOnNode(
//TODO (yastij) : compute average predicate restrictiveness to export it as Prometheus metric
if predicate, exist := predicateFuncs[predicateKey]; exist {
if eCacheAvailable {
fit, reasons, err = ecache.RunPredicate(predicate, predicateKey, pod, metaToUse, nodeInfoToUse, equivCacheInfo, cache)
fit, reasons, err = nodeCache.RunPredicate(predicate, predicateKey, predicateID, pod, metaToUse, nodeInfoToUse, equivClass)
} else {
fit, reasons, err = predicate(pod, metaToUse, nodeInfoToUse)
}
@ -524,8 +599,8 @@ func podFitsOnNode(
failedPredicates = append(failedPredicates, reasons...)
// if alwaysCheckAllPredicates is false, short circuit all predicates when one predicate fails.
if !alwaysCheckAllPredicates {
glog.V(5).Infoln("since alwaysCheckAllPredicates has not been set, the predicate" +
"evaluation is short circuited and there are chances" +
klog.V(5).Infoln("since alwaysCheckAllPredicates has not been set, the predicate " +
"evaluation is short circuited and there are chances " +
"of other predicates failing as well.")
break
}
@ -578,37 +653,38 @@ func PrioritizeNodes(
results := make([]schedulerapi.HostPriorityList, len(priorityConfigs), len(priorityConfigs))
for i, priorityConfig := range priorityConfigs {
if priorityConfig.Function != nil {
// DEPRECATED
wg.Add(1)
go func(index int, config algorithm.PriorityConfig) {
defer wg.Done()
var err error
results[index], err = config.Function(pod, nodeNameToInfo, nodes)
if err != nil {
appendError(err)
}
}(i, priorityConfig)
} else {
// DEPRECATED: we can remove this when all priorityConfigs implement the
// Map-Reduce pattern.
workqueue.ParallelizeUntil(context.TODO(), 16, len(priorityConfigs), func(i int) {
priorityConfig := priorityConfigs[i]
if priorityConfig.Function == nil {
results[i] = make(schedulerapi.HostPriorityList, len(nodes))
return
}
}
processNode := func(index int) {
nodeInfo := nodeNameToInfo[nodes[index].Name]
var err error
for i := range priorityConfigs {
if priorityConfigs[i].Function != nil {
results[i], err = priorityConfig.Function(pod, nodeNameToInfo, nodes)
if err != nil {
appendError(err)
}
})
workqueue.ParallelizeUntil(context.TODO(), 16, len(nodes), func(index int) {
nodeInfo := nodeNameToInfo[nodes[index].Name]
for i, priorityConfig := range priorityConfigs {
if priorityConfig.Function != nil {
continue
}
var err error
results[i][index], err = priorityConfigs[i].Map(pod, meta, nodeInfo)
if err != nil {
appendError(err)
return
results[i][index].Host = nodes[index].Name
}
}
}
workqueue.Parallelize(16, len(nodes), processNode)
})
for i, priorityConfig := range priorityConfigs {
if priorityConfig.Reduce == nil {
continue
@ -619,9 +695,9 @@ func PrioritizeNodes(
if err := config.Reduce(pod, meta, nodeNameToInfo, results[index]); err != nil {
appendError(err)
}
if glog.V(10) {
if klog.V(10) {
for _, hostPriority := range results[index] {
glog.Infof("%v -> %v: %v, Score: (%d)", pod.Name, hostPriority.Host, config.Name, hostPriority.Score)
klog.Infof("%v -> %v: %v, Score: (%d)", util.GetPodFullName(pod), hostPriority.Host, config.Name, hostPriority.Score)
}
}
}(i, priorityConfig)
@ -659,6 +735,9 @@ func PrioritizeNodes(
mu.Lock()
for i := range *prioritizedList {
host, score := (*prioritizedList)[i].Host, (*prioritizedList)[i].Score
if klog.V(10) {
klog.Infof("%v -> %v: %v, Score: (%d)", util.GetPodFullName(pod), host, ext.Name(), score)
}
combinedScores[host] += score * weight
}
mu.Unlock()
@ -671,9 +750,9 @@ func PrioritizeNodes(
}
}
if glog.V(10) {
if klog.V(10) {
for i := range result {
glog.V(10).Infof("Host %s => Score %d", result[i].Host, result[i].Score)
klog.Infof("Host %s => Score %d", result[i].Host, result[i].Score)
}
}
return result, nil
@ -802,7 +881,7 @@ func pickOneNodeForPreemption(nodesToVictims map[*v1.Node]*schedulerapi.Victims)
if lenNodes2 > 0 {
return minNodes2[0]
}
glog.Errorf("Error in logic of node scoring for preemption. We should never reach here!")
klog.Errorf("Error in logic of node scoring for preemption. We should never reach here!")
return nil
}
@ -813,10 +892,9 @@ func selectNodesForPreemption(pod *v1.Pod,
potentialNodes []*v1.Node,
predicates map[string]algorithm.FitPredicate,
metadataProducer algorithm.PredicateMetadataProducer,
queue SchedulingQueue,
queue internalqueue.SchedulingQueue,
pdbs []*policy.PodDisruptionBudget,
) (map[*v1.Node]*schedulerapi.Victims, error) {
nodeToVictims := map[*v1.Node]*schedulerapi.Victims{}
var resultLock sync.Mutex
@ -839,7 +917,7 @@ func selectNodesForPreemption(pod *v1.Pod,
resultLock.Unlock()
}
}
workqueue.Parallelize(16, len(potentialNodes), checkNode)
workqueue.ParallelizeUntil(context.TODO(), 16, len(potentialNodes), checkNode)
return nodeToVictims, nil
}
@ -902,9 +980,12 @@ func selectVictimsOnNode(
meta algorithm.PredicateMetadata,
nodeInfo *schedulercache.NodeInfo,
fitPredicates map[string]algorithm.FitPredicate,
queue SchedulingQueue,
queue internalqueue.SchedulingQueue,
pdbs []*policy.PodDisruptionBudget,
) ([]*v1.Pod, int, bool) {
if nodeInfo == nil {
return nil, 0, false
}
potentialVictims := util.SortableList{CompFunc: util.HigherPriorityPod}
nodeInfoCopy := nodeInfo.Clone()
@ -935,9 +1016,9 @@ func selectVictimsOnNode(
// that we should check is if the "pod" is failing to schedule due to pod affinity
// failure.
// TODO(bsalamat): Consider checking affinity to lower priority pods if feasible with reasonable performance.
if fits, _, err := podFitsOnNode(pod, meta, nodeInfoCopy, fitPredicates, nil, nil, queue, false, nil); !fits {
if fits, _, err := podFitsOnNode(pod, meta, nodeInfoCopy, fitPredicates, nil, queue, false, nil); !fits {
if err != nil {
glog.Warningf("Encountered error while selecting victims on node %v: %v", nodeInfo.Node().Name, err)
klog.Warningf("Encountered error while selecting victims on node %v: %v", nodeInfo.Node().Name, err)
}
return nil, 0, false
}
@ -949,11 +1030,11 @@ func selectVictimsOnNode(
violatingVictims, nonViolatingVictims := filterPodsWithPDBViolation(potentialVictims.Items, pdbs)
reprievePod := func(p *v1.Pod) bool {
addPod(p)
fits, _, _ := podFitsOnNode(pod, meta, nodeInfoCopy, fitPredicates, nil, nil, queue, false, nil)
fits, _, _ := podFitsOnNode(pod, meta, nodeInfoCopy, fitPredicates, nil, queue, false, nil)
if !fits {
removePod(p)
victims = append(victims, p)
glog.V(5).Infof("Pod %v is a potential preemption victim on node %v.", p.Name, nodeInfo.Node().Name)
klog.V(5).Infof("Pod %v is a potential preemption victim on node %v.", p.Name, nodeInfo.Node().Name)
}
return fits
}
@ -971,7 +1052,7 @@ func selectVictimsOnNode(
// nodesWherePreemptionMightHelp returns a list of nodes with failed predicates
// that may be satisfied by removing pods from the node.
func nodesWherePreemptionMightHelp(pod *v1.Pod, nodes []*v1.Node, failedPredicatesMap FailedPredicateMap) []*v1.Node {
func nodesWherePreemptionMightHelp(nodes []*v1.Node, failedPredicatesMap FailedPredicateMap) []*v1.Node {
potentialNodes := []*v1.Node{}
for _, node := range nodes {
unresolvableReasonExist := false
@ -985,11 +1066,19 @@ func nodesWherePreemptionMightHelp(pod *v1.Pod, nodes []*v1.Node, failedPredicat
switch failedPredicate {
case
predicates.ErrNodeSelectorNotMatch,
predicates.ErrPodAffinityRulesNotMatch,
predicates.ErrPodNotMatchHostName,
predicates.ErrTaintsTolerationsNotMatch,
predicates.ErrNodeLabelPresenceViolated,
// Node conditions won't change when scheduler simulates removal of preemption victims.
// So, it is pointless to try nodes that have not been able to host the pod due to node
// conditions. These include ErrNodeNotReady, ErrNodeUnderPIDPressure, ErrNodeUnderMemoryPressure, ....
predicates.ErrNodeNotReady,
predicates.ErrNodeNetworkUnavailable,
predicates.ErrNodeUnderDiskPressure,
predicates.ErrNodeUnderPIDPressure,
predicates.ErrNodeUnderMemoryPressure,
predicates.ErrNodeOutOfDisk,
predicates.ErrNodeUnschedulable,
predicates.ErrNodeUnknownCondition,
predicates.ErrVolumeZoneConflict,
@ -997,11 +1086,10 @@ func nodesWherePreemptionMightHelp(pod *v1.Pod, nodes []*v1.Node, failedPredicat
predicates.ErrVolumeBindConflict:
unresolvableReasonExist = true
break
// TODO(bsalamat): Please add affinity failure cases once we have specific affinity failure errors.
}
}
if !found || !unresolvableReasonExist {
glog.V(3).Infof("Node %v is a potential node for preemption.", node.Name)
klog.V(3).Infof("Node %v is a potential node for preemption.", node.Name)
potentialNodes = append(potentialNodes, node)
}
}
@ -1057,9 +1145,9 @@ func podPassesBasicChecks(pod *v1.Pod, pvcLister corelisters.PersistentVolumeCla
// NewGenericScheduler creates a genericScheduler object.
func NewGenericScheduler(
cache schedulercache.Cache,
eCache *EquivalenceCache,
podQueue SchedulingQueue,
cache schedulerinternalcache.Cache,
eCache *equivalence.Cache,
podQueue internalqueue.SchedulingQueue,
predicates map[string]algorithm.FitPredicate,
predicateMetaProducer algorithm.PredicateMetadataProducer,
prioritizers []algorithm.PriorityConfig,
@ -1067,8 +1155,10 @@ func NewGenericScheduler(
extenders []algorithm.SchedulerExtender,
volumeBinder *volumebinder.VolumeBinder,
pvcLister corelisters.PersistentVolumeClaimLister,
pdbLister algorithm.PDBLister,
alwaysCheckAllPredicates bool,
disablePreemption bool,
percentageOfNodesToScore int32,
) algorithm.ScheduleAlgorithm {
return &genericScheduler{
cache: cache,
@ -1082,7 +1172,9 @@ func NewGenericScheduler(
cachedNodeInfoMap: make(map[string]*schedulercache.NodeInfo),
volumeBinder: volumeBinder,
pvcLister: pvcLister,
pdbLister: pdbLister,
alwaysCheckAllPredicates: alwaysCheckAllPredicates,
disablePreemption: disablePreemption,
percentageOfNodesToScore: percentageOfNodesToScore,
}
}

File diff suppressed because it is too large Load Diff

View File

@ -1,128 +1,88 @@
package(default_visibility = ["//visibility:public"])
load(
"@io_bazel_rules_go//go:def.bzl",
"go_library",
"go_test",
)
load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
go_library(
name = "go_default_library",
srcs = [
"cache_comparer.go",
"factory.go",
"plugins.go",
] + select({
"@io_bazel_rules_go//go/platform:android": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:darwin": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:dragonfly": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:freebsd": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:linux": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:nacl": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:netbsd": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:openbsd": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:plan9": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:solaris": [
"signal.go",
],
"@io_bazel_rules_go//go/platform:windows": [
"signal_windows.go",
],
"//conditions:default": [],
}),
"signal.go",
"signal_windows.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/factory",
visibility = ["//visibility:public"],
deps = [
"//pkg/api/v1/pod:go_default_library",
"//pkg/apis/core/helper:go_default_library",
"//pkg/features:go_default_library",
"//pkg/kubelet/apis:go_default_library",
"//pkg/scheduler:go_default_library",
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/predicates:go_default_library",
"//pkg/scheduler/algorithm/priorities:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/api/validation:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/core:go_default_library",
"//pkg/scheduler/core/equivalence:go_default_library",
"//pkg/scheduler/internal/cache:go_default_library",
"//pkg/scheduler/internal/cache/debugger:go_default_library",
"//pkg/scheduler/internal/queue:go_default_library",
"//pkg/scheduler/util:go_default_library",
"//pkg/scheduler/volumebinder:go_default_library",
"//vendor/github.com/golang/glog:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/policy/v1beta1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/api/errors:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/fields:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/types:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/runtime:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//vendor/k8s.io/client-go/informers/apps/v1beta1:go_default_library",
"//vendor/k8s.io/client-go/informers/core/v1:go_default_library",
"//vendor/k8s.io/client-go/informers/extensions/v1beta1:go_default_library",
"//vendor/k8s.io/client-go/informers/policy/v1beta1:go_default_library",
"//vendor/k8s.io/client-go/informers/storage/v1:go_default_library",
"//vendor/k8s.io/client-go/kubernetes:go_default_library",
"//vendor/k8s.io/client-go/listers/apps/v1beta1:go_default_library",
"//vendor/k8s.io/client-go/listers/core/v1:go_default_library",
"//vendor/k8s.io/client-go/listers/extensions/v1beta1:go_default_library",
"//vendor/k8s.io/client-go/listers/policy/v1beta1:go_default_library",
"//vendor/k8s.io/client-go/listers/storage/v1:go_default_library",
"//vendor/k8s.io/client-go/tools/cache:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/api/storage/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/errors:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/fields:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/types:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/wait:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/client-go/informers/apps/v1:go_default_library",
"//staging/src/k8s.io/client-go/informers/core/v1:go_default_library",
"//staging/src/k8s.io/client-go/informers/policy/v1beta1:go_default_library",
"//staging/src/k8s.io/client-go/informers/storage/v1:go_default_library",
"//staging/src/k8s.io/client-go/kubernetes:go_default_library",
"//staging/src/k8s.io/client-go/listers/apps/v1:go_default_library",
"//staging/src/k8s.io/client-go/listers/core/v1:go_default_library",
"//staging/src/k8s.io/client-go/listers/policy/v1beta1:go_default_library",
"//staging/src/k8s.io/client-go/listers/storage/v1:go_default_library",
"//staging/src/k8s.io/client-go/tools/cache:go_default_library",
"//staging/src/k8s.io/client-go/tools/record:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = [
"cache_comparer_test.go",
"factory_test.go",
"plugins_test.go",
],
embed = [":go_default_library"],
deps = [
"//pkg/api/testing:go_default_library",
"//pkg/scheduler:go_default_library",
"//pkg/scheduler/algorithm:go_default_library",
"//pkg/scheduler/algorithm/priorities:go_default_library",
"//pkg/scheduler/api:go_default_library",
"//pkg/scheduler/api/latest:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/core:go_default_library",
"//pkg/scheduler/internal/cache/fake:go_default_library",
"//pkg/scheduler/internal/queue:go_default_library",
"//pkg/scheduler/testing:go_default_library",
"//pkg/scheduler/util:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/client-go/informers:go_default_library",
"//staging/src/k8s.io/client-go/kubernetes:go_default_library",
"//staging/src/k8s.io/client-go/kubernetes/fake:go_default_library",
"//staging/src/k8s.io/client-go/kubernetes/typed/core/v1/fake:go_default_library",
"//staging/src/k8s.io/client-go/testing:go_default_library",
"//staging/src/k8s.io/client-go/tools/cache:go_default_library",
"//vendor/github.com/stretchr/testify/assert:go_default_library",
"//vendor/k8s.io/api/core/v1:go_default_library",
"//vendor/k8s.io/api/policy/v1beta1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/runtime/schema:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/types:go_default_library",
"//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//vendor/k8s.io/client-go/informers:go_default_library",
"//vendor/k8s.io/client-go/kubernetes:go_default_library",
"//vendor/k8s.io/client-go/rest:go_default_library",
"//vendor/k8s.io/client-go/tools/cache:go_default_library",
"//vendor/k8s.io/client-go/util/testing:go_default_library",
],
)
@ -137,4 +97,5 @@ filegroup(
name = "all-srcs",
srcs = [":package-srcs"],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

View File

@ -1,228 +0,0 @@
/*
Copyright 2018 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 factory
import (
"reflect"
"testing"
"k8s.io/api/core/v1"
policy "k8s.io/api/policy/v1beta1"
"k8s.io/apimachinery/pkg/types"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
)
func TestCompareNodes(t *testing.T) {
compare := compareStrategy{}
tests := []struct {
actual []string
cached []string
missing []string
redundant []string
}{
{
actual: []string{"foo", "bar"},
cached: []string{"bar", "foo", "foobar"},
missing: []string{},
redundant: []string{"foobar"},
},
{
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foo"},
missing: []string{"foobar"},
redundant: []string{},
},
{
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foobar", "foo"},
missing: []string{},
redundant: []string{},
},
}
for _, test := range tests {
nodes := []*v1.Node{}
for _, nodeName := range test.actual {
node := &v1.Node{}
node.Name = nodeName
nodes = append(nodes, node)
}
nodeInfo := make(map[string]*schedulercache.NodeInfo)
for _, nodeName := range test.cached {
nodeInfo[nodeName] = &schedulercache.NodeInfo{}
}
m, r := compare.CompareNodes(nodes, nodeInfo)
if !reflect.DeepEqual(m, test.missing) {
t.Errorf("missing expected to be %s; got %s", test.missing, m)
}
if !reflect.DeepEqual(r, test.redundant) {
t.Errorf("redundant expected to be %s; got %s", test.redundant, r)
}
}
}
func TestComparePods(t *testing.T) {
compare := compareStrategy{}
tests := []struct {
actual []string
cached []string
queued []string
missing []string
redundant []string
}{
{
actual: []string{"foo", "bar"},
cached: []string{"bar", "foo", "foobar"},
queued: []string{},
missing: []string{},
redundant: []string{"foobar"},
},
{
actual: []string{"foo", "bar"},
cached: []string{"foo", "foobar"},
queued: []string{"bar"},
missing: []string{},
redundant: []string{"foobar"},
},
{
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foo"},
queued: []string{},
missing: []string{"foobar"},
redundant: []string{},
},
{
actual: []string{"foo", "bar", "foobar"},
cached: []string{"foo"},
queued: []string{"bar"},
missing: []string{"foobar"},
redundant: []string{},
},
{
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foobar", "foo"},
queued: []string{},
missing: []string{},
redundant: []string{},
},
{
actual: []string{"foo", "bar", "foobar"},
cached: []string{"foobar", "foo"},
queued: []string{"bar"},
missing: []string{},
redundant: []string{},
},
}
for _, test := range tests {
pods := []*v1.Pod{}
for _, uid := range test.actual {
pod := &v1.Pod{}
pod.UID = types.UID(uid)
pods = append(pods, pod)
}
queuedPods := []*v1.Pod{}
for _, uid := range test.queued {
pod := &v1.Pod{}
pod.UID = types.UID(uid)
queuedPods = append(queuedPods, pod)
}
nodeInfo := make(map[string]*schedulercache.NodeInfo)
for _, uid := range test.cached {
pod := &v1.Pod{}
pod.UID = types.UID(uid)
pod.Namespace = "ns"
pod.Name = uid
nodeInfo[uid] = schedulercache.NewNodeInfo(pod)
}
m, r := compare.ComparePods(pods, queuedPods, nodeInfo)
if !reflect.DeepEqual(m, test.missing) {
t.Errorf("missing expected to be %s; got %s", test.missing, m)
}
if !reflect.DeepEqual(r, test.redundant) {
t.Errorf("redundant expected to be %s; got %s", test.redundant, r)
}
}
}
func TestComparePdbs(t *testing.T) {
compare := compareStrategy{}
tests := []struct {
actual []string
cached []string
missing []string
redundant []string
}{
{
actual: []string{"foo", "bar"},
cached: []string{"bar", "foo", "foobar"},
missing: []string{},
redundant: []string{"foobar"},
},
{
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foo"},
missing: []string{"foobar"},
redundant: []string{},
},
{
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foobar", "foo"},
missing: []string{},
redundant: []string{},
},
}
for _, test := range tests {
pdbs := []*policy.PodDisruptionBudget{}
for _, uid := range test.actual {
pdb := &policy.PodDisruptionBudget{}
pdb.UID = types.UID(uid)
pdbs = append(pdbs, pdb)
}
cache := make(map[string]*policy.PodDisruptionBudget)
for _, uid := range test.cached {
pdb := &policy.PodDisruptionBudget{}
pdb.UID = types.UID(uid)
cache[uid] = pdb
}
m, r := compare.ComparePdbs(pdbs, cache)
if !reflect.DeepEqual(m, test.missing) {
t.Errorf("missing expected to be %s; got %s", test.missing, m)
}
if !reflect.DeepEqual(r, test.redundant) {
t.Errorf("redundant expected to be %s; got %s", test.redundant, r)
}
}
}

File diff suppressed because it is too large Load Diff

View File

@ -19,8 +19,6 @@ package factory
import (
"errors"
"fmt"
"net/http"
"net/http/httptest"
"reflect"
"testing"
"time"
@ -28,20 +26,20 @@ import (
"k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/runtime"
"k8s.io/apimachinery/pkg/runtime/schema"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/client-go/informers"
clientset "k8s.io/client-go/kubernetes"
restclient "k8s.io/client-go/rest"
"k8s.io/client-go/kubernetes/fake"
fakeV1 "k8s.io/client-go/kubernetes/typed/core/v1/fake"
clienttesting "k8s.io/client-go/testing"
"k8s.io/client-go/tools/cache"
utiltesting "k8s.io/client-go/util/testing"
apitesting "k8s.io/kubernetes/pkg/api/testing"
"k8s.io/kubernetes/pkg/scheduler"
"k8s.io/kubernetes/pkg/scheduler/algorithm"
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
latestschedulerapi "k8s.io/kubernetes/pkg/scheduler/api/latest"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
"k8s.io/kubernetes/pkg/scheduler/core"
fakecache "k8s.io/kubernetes/pkg/scheduler/internal/cache/fake"
internalqueue "k8s.io/kubernetes/pkg/scheduler/internal/queue"
schedulertesting "k8s.io/kubernetes/pkg/scheduler/testing"
"k8s.io/kubernetes/pkg/scheduler/util"
)
@ -49,18 +47,14 @@ import (
const (
enableEquivalenceCache = true
disablePodPreemption = false
bindTimeoutSeconds = 600
)
func TestCreate(t *testing.T) {
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight)
client := fake.NewSimpleClientset()
stopCh := make(chan struct{})
defer close(stopCh)
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight, stopCh)
factory.Create()
}
@ -70,15 +64,10 @@ func TestCreateFromConfig(t *testing.T) {
var configData []byte
var policy schedulerapi.Policy
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight)
client := fake.NewSimpleClientset()
stopCh := make(chan struct{})
defer close(stopCh)
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight, stopCh)
// Pre-register some predicate and priority functions
RegisterFitPredicate("PredicateOne", PredicateOne)
@ -115,15 +104,10 @@ func TestCreateFromConfigWithHardPodAffinitySymmetricWeight(t *testing.T) {
var configData []byte
var policy schedulerapi.Policy
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight)
client := fake.NewSimpleClientset()
stopCh := make(chan struct{})
defer close(stopCh)
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight, stopCh)
// Pre-register some predicate and priority functions
RegisterFitPredicate("PredicateOne", PredicateOne)
@ -161,15 +145,10 @@ func TestCreateFromEmptyConfig(t *testing.T) {
var configData []byte
var policy schedulerapi.Policy
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight)
client := fake.NewSimpleClientset()
stopCh := make(chan struct{})
defer close(stopCh)
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight, stopCh)
configData = []byte(`{}`)
if err := runtime.DecodeInto(latestschedulerapi.Codec, configData, &policy); err != nil {
@ -183,15 +162,10 @@ func TestCreateFromEmptyConfig(t *testing.T) {
// predicate/priority.
// The predicate/priority from DefaultProvider will be used.
func TestCreateFromConfigWithUnspecifiedPredicatesOrPriorities(t *testing.T) {
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight)
client := fake.NewSimpleClientset()
stopCh := make(chan struct{})
defer close(stopCh)
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight, stopCh)
RegisterFitPredicate("PredicateOne", PredicateOne)
RegisterPriorityFunction("PriorityOne", PriorityOne, 1)
@ -223,15 +197,10 @@ func TestCreateFromConfigWithUnspecifiedPredicatesOrPriorities(t *testing.T) {
// predicate/priority.
// Empty predicate/priority sets will be used.
func TestCreateFromConfigWithEmptyPredicatesOrPriorities(t *testing.T) {
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight)
client := fake.NewSimpleClientset()
stopCh := make(chan struct{})
defer close(stopCh)
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight, stopCh)
RegisterFitPredicate("PredicateOne", PredicateOne)
RegisterPriorityFunction("PriorityOne", PriorityOne, 1)
@ -282,24 +251,16 @@ func TestDefaultErrorFunc(t *testing.T) {
ObjectMeta: metav1.ObjectMeta{Name: "foo", Namespace: "bar"},
Spec: apitesting.V1DeepEqualSafePodSpec(),
}
handler := utiltesting.FakeHandler{
StatusCode: 200,
ResponseBody: runtime.EncodeOrDie(schedulertesting.Test.Codec(), testPod),
T: t,
}
mux := http.NewServeMux()
// FakeHandler mustn't be sent requests other than the one you want to test.
mux.Handle(schedulertesting.Test.ResourcePath(string(v1.ResourcePods), "bar", "foo"), &handler)
server := httptest.NewServer(mux)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight)
queue := &core.FIFO{FIFO: cache.NewFIFO(cache.MetaNamespaceKeyFunc)}
client := fake.NewSimpleClientset(&v1.PodList{Items: []v1.Pod{*testPod}})
stopCh := make(chan struct{})
defer close(stopCh)
factory := newConfigFactory(client, v1.DefaultHardPodAffinitySymmetricWeight, stopCh)
queue := &internalqueue.FIFO{FIFO: cache.NewFIFO(cache.MetaNamespaceKeyFunc)}
podBackoff := util.CreatePodBackoff(1*time.Millisecond, 1*time.Second)
errFunc := factory.MakeDefaultErrorFunc(podBackoff, queue)
errFunc(testPod, nil)
for {
// This is a terrible way to do this but I plan on replacing this
// whole error handling system in the future. The test will time
@ -309,7 +270,27 @@ func TestDefaultErrorFunc(t *testing.T) {
if !exists {
continue
}
handler.ValidateRequest(t, schedulertesting.Test.ResourcePath(string(v1.ResourcePods), "bar", "foo"), "GET", nil)
requestReceived := false
actions := client.Actions()
for _, a := range actions {
if a.GetVerb() == "get" {
getAction, ok := a.(clienttesting.GetAction)
if !ok {
t.Errorf("Can't cast action object to GetAction interface")
break
}
name := getAction.GetName()
ns := a.GetNamespace()
if name != "foo" || ns != "bar" {
t.Errorf("Expected name %s namespace %s, got %s %s",
"foo", "bar", name, ns)
}
requestReceived = true
}
}
if !requestReceived {
t.Errorf("Get pod request not received")
}
if e, a := testPod, got; !reflect.DeepEqual(e, a) {
t.Errorf("Expected %v, got %v", e, a)
}
@ -331,64 +312,81 @@ func TestNodeEnumerator(t *testing.T) {
t.Fatalf("expected %v, got %v", e, a)
}
for i := range testList.Items {
gotObj := me.Get(i)
if e, a := testList.Items[i].Name, gotObj.(*v1.Node).Name; e != a {
t.Errorf("Expected %v, got %v", e, a)
}
if e, a := &testList.Items[i], gotObj; !reflect.DeepEqual(e, a) {
t.Errorf("Expected %#v, got %v#", e, a)
}
t.Run(fmt.Sprintf("node enumerator/%v", i), func(t *testing.T) {
gotObj := me.Get(i)
if e, a := testList.Items[i].Name, gotObj.(*v1.Node).Name; e != a {
t.Errorf("Expected %v, got %v", e, a)
}
if e, a := &testList.Items[i], gotObj; !reflect.DeepEqual(e, a) {
t.Errorf("Expected %#v, got %v#", e, a)
}
})
}
}
func TestBind(t *testing.T) {
table := []struct {
name string
binding *v1.Binding
}{
{binding: &v1.Binding{
ObjectMeta: metav1.ObjectMeta{
Namespace: metav1.NamespaceDefault,
Name: "foo",
{
name: "binding can bind and validate request",
binding: &v1.Binding{
ObjectMeta: metav1.ObjectMeta{
Namespace: metav1.NamespaceDefault,
Name: "foo",
},
Target: v1.ObjectReference{
Name: "foohost.kubernetes.mydomain.com",
},
},
Target: v1.ObjectReference{
Name: "foohost.kubernetes.mydomain.com",
},
}},
},
}
for _, item := range table {
handler := utiltesting.FakeHandler{
StatusCode: 200,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
b := binder{client}
for _, test := range table {
t.Run(test.name, func(t *testing.T) {
testBind(test.binding, t)
})
}
}
if err := b.Bind(item.binding); err != nil {
t.Errorf("Unexpected error: %v", err)
continue
}
expectedBody := runtime.EncodeOrDie(schedulertesting.Test.Codec(), item.binding)
handler.ValidateRequest(t,
schedulertesting.Test.SubResourcePath(string(v1.ResourcePods), metav1.NamespaceDefault, "foo", "binding"),
"POST", &expectedBody)
func testBind(binding *v1.Binding, t *testing.T) {
testPod := &v1.Pod{
ObjectMeta: metav1.ObjectMeta{Name: binding.GetName(), Namespace: metav1.NamespaceDefault},
Spec: apitesting.V1DeepEqualSafePodSpec(),
}
client := fake.NewSimpleClientset(&v1.PodList{Items: []v1.Pod{*testPod}})
b := binder{client}
if err := b.Bind(binding); err != nil {
t.Errorf("Unexpected error: %v", err)
return
}
pod := client.CoreV1().Pods(metav1.NamespaceDefault).(*fakeV1.FakePods)
bind, err := pod.GetBinding(binding.GetName())
if err != nil {
t.Fatalf("Unexpected error: %v", err)
return
}
expectedBody := runtime.EncodeOrDie(schedulertesting.Test.Codec(), binding)
bind.APIVersion = ""
bind.Kind = ""
body := runtime.EncodeOrDie(schedulertesting.Test.Codec(), bind)
if expectedBody != body {
t.Errorf("Expected body %s, Got %s", expectedBody, body)
}
}
func TestInvalidHardPodAffinitySymmetricWeight(t *testing.T) {
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
client := fake.NewSimpleClientset()
// factory of "default-scheduler"
factory := newConfigFactory(client, -1)
stopCh := make(chan struct{})
factory := newConfigFactory(client, -1, stopCh)
defer close(stopCh)
_, err := factory.Create()
if err == nil {
t.Errorf("expected err: invalid hardPodAffinitySymmetricWeight, got nothing")
@ -396,48 +394,49 @@ func TestInvalidHardPodAffinitySymmetricWeight(t *testing.T) {
}
func TestInvalidFactoryArgs(t *testing.T) {
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := clientset.NewForConfigOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: &schema.GroupVersion{Group: "", Version: "v1"}}})
client := fake.NewSimpleClientset()
testCases := []struct {
name string
hardPodAffinitySymmetricWeight int32
expectErr string
}{
{
name: "symmetric weight below range",
hardPodAffinitySymmetricWeight: -1,
expectErr: "invalid hardPodAffinitySymmetricWeight: -1, must be in the range 0-100",
},
{
name: "symmetric weight above range",
hardPodAffinitySymmetricWeight: 101,
expectErr: "invalid hardPodAffinitySymmetricWeight: 101, must be in the range 0-100",
},
}
for _, test := range testCases {
factory := newConfigFactory(client, test.hardPodAffinitySymmetricWeight)
_, err := factory.Create()
if err == nil {
t.Errorf("expected err: %s, got nothing", test.expectErr)
}
t.Run(test.name, func(t *testing.T) {
stopCh := make(chan struct{})
factory := newConfigFactory(client, test.hardPodAffinitySymmetricWeight, stopCh)
defer close(stopCh)
_, err := factory.Create()
if err == nil {
t.Errorf("expected err: %s, got nothing", test.expectErr)
}
})
}
}
func TestSkipPodUpdate(t *testing.T) {
for _, test := range []struct {
table := []struct {
pod *v1.Pod
isAssumedPodFunc func(*v1.Pod) bool
getPodFunc func(*v1.Pod) *v1.Pod
expected bool
name string
}{
// Non-assumed pod should not be skipped.
{
name: "Non-assumed pod",
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{
Name: "pod-0",
@ -453,9 +452,8 @@ func TestSkipPodUpdate(t *testing.T) {
},
expected: false,
},
// Pod update (with changes on ResourceVersion, Spec.NodeName and/or
// Annotations) for an already assumed pod should be skipped.
{
name: "with changes on ResourceVersion, Spec.NodeName and/or Annotations",
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{
Name: "pod-0",
@ -483,9 +481,8 @@ func TestSkipPodUpdate(t *testing.T) {
},
expected: true,
},
// Pod update (with changes on Labels) for an already assumed pod
// should not be skipped.
{
name: "with changes on Labels",
pod: &v1.Pod{
ObjectMeta: metav1.ObjectMeta{
Name: "pod-0",
@ -505,23 +502,26 @@ func TestSkipPodUpdate(t *testing.T) {
},
expected: false,
},
} {
c := &configFactory{
schedulerCache: &schedulertesting.FakeCache{
IsAssumedPodFunc: test.isAssumedPodFunc,
GetPodFunc: test.getPodFunc,
},
}
got := c.skipPodUpdate(test.pod)
if got != test.expected {
t.Errorf("skipPodUpdate() = %t, expected = %t", got, test.expected)
}
}
for _, test := range table {
t.Run(test.name, func(t *testing.T) {
c := &configFactory{
schedulerCache: &fakecache.Cache{
IsAssumedPodFunc: test.isAssumedPodFunc,
GetPodFunc: test.getPodFunc,
},
}
got := c.skipPodUpdate(test.pod)
if got != test.expected {
t.Errorf("skipPodUpdate() = %t, expected = %t", got, test.expected)
}
})
}
}
func newConfigFactory(client *clientset.Clientset, hardPodAffinitySymmetricWeight int32) scheduler.Configurator {
func newConfigFactory(client clientset.Interface, hardPodAffinitySymmetricWeight int32, stopCh <-chan struct{}) Configurator {
informerFactory := informers.NewSharedInformerFactory(client, 0)
return NewConfigFactory(
return NewConfigFactory(&ConfigFactoryArgs{
v1.DefaultSchedulerName,
client,
informerFactory.Core().V1().Nodes(),
@ -529,15 +529,18 @@ func newConfigFactory(client *clientset.Clientset, hardPodAffinitySymmetricWeigh
informerFactory.Core().V1().PersistentVolumes(),
informerFactory.Core().V1().PersistentVolumeClaims(),
informerFactory.Core().V1().ReplicationControllers(),
informerFactory.Extensions().V1beta1().ReplicaSets(),
informerFactory.Apps().V1beta1().StatefulSets(),
informerFactory.Apps().V1().ReplicaSets(),
informerFactory.Apps().V1().StatefulSets(),
informerFactory.Core().V1().Services(),
informerFactory.Policy().V1beta1().PodDisruptionBudgets(),
informerFactory.Storage().V1().StorageClasses(),
hardPodAffinitySymmetricWeight,
enableEquivalenceCache,
disablePodPreemption,
)
schedulerapi.DefaultPercentageOfNodesToScore,
bindTimeoutSeconds,
stopCh,
})
}
type fakeExtender struct {
@ -546,6 +549,10 @@ type fakeExtender struct {
ignorable bool
}
func (f *fakeExtender) Name() string {
return "fakeExtender"
}
func (f *fakeExtender) IsIgnorable() bool {
return f.ignorable
}
@ -593,24 +600,22 @@ func (f *fakeExtender) IsInterested(pod *v1.Pod) bool {
}
func TestGetBinderFunc(t *testing.T) {
for _, test := range []struct {
podName string
extenders []algorithm.SchedulerExtender
table := []struct {
podName string
extenders []algorithm.SchedulerExtender
expectedBinderType string
name string
}{
// Expect to return the default binder because the extender is not a
// binder, even though it's interested in the pod.
{
name: "the extender is not a binder",
podName: "pod0",
extenders: []algorithm.SchedulerExtender{
&fakeExtender{isBinder: false, interestedPodName: "pod0"},
},
expectedBinderType: "*factory.binder",
},
// Expect to return the fake binder because one of the extenders is a
// binder and it's interested in the pod.
{
name: "one of the extenders is a binder and interested in pod",
podName: "pod0",
extenders: []algorithm.SchedulerExtender{
&fakeExtender{isBinder: false, interestedPodName: "pod0"},
@ -618,9 +623,8 @@ func TestGetBinderFunc(t *testing.T) {
},
expectedBinderType: "*factory.fakeExtender",
},
// Expect to return the default binder because one of the extenders is
// a binder but the binder is not interested in the pod.
{
name: "one of the extenders is a binder, but not interested in pod",
podName: "pod1",
extenders: []algorithm.SchedulerExtender{
&fakeExtender{isBinder: false, interestedPodName: "pod1"},
@ -628,20 +632,28 @@ func TestGetBinderFunc(t *testing.T) {
},
expectedBinderType: "*factory.binder",
},
} {
pod := &v1.Pod{
ObjectMeta: metav1.ObjectMeta{
Name: test.podName,
},
}
}
f := &configFactory{}
binderFunc := f.getBinderFunc(test.extenders)
binder := binderFunc(pod)
binderType := fmt.Sprintf("%s", reflect.TypeOf(binder))
if binderType != test.expectedBinderType {
t.Errorf("Expected binder %q but got %q", test.expectedBinderType, binderType)
}
for _, test := range table {
t.Run(test.name, func(t *testing.T) {
testGetBinderFunc(test.expectedBinderType, test.podName, test.extenders, t)
})
}
}
func testGetBinderFunc(expectedBinderType, podName string, extenders []algorithm.SchedulerExtender, t *testing.T) {
pod := &v1.Pod{
ObjectMeta: metav1.ObjectMeta{
Name: podName,
},
}
f := &configFactory{}
binderFunc := f.getBinderFunc(extenders)
binder := binderFunc(pod)
binderType := fmt.Sprintf("%s", reflect.TypeOf(binder))
if binderType != expectedBinderType {
t.Errorf("Expected binder %q but got %q", expectedBinderType, binderType)
}
}

View File

@ -30,7 +30,7 @@ import (
schedulerapi "k8s.io/kubernetes/pkg/scheduler/api"
"k8s.io/kubernetes/pkg/scheduler/volumebinder"
"github.com/golang/glog"
"k8s.io/klog"
)
// PluginFactoryArgs are passed to all plugin factory functions.
@ -41,6 +41,7 @@ type PluginFactoryArgs struct {
ReplicaSetLister algorithm.ReplicaSetLister
StatefulSetLister algorithm.StatefulSetLister
NodeLister algorithm.NodeLister
PDBLister algorithm.PDBLister
NodeInfo predicates.NodeInfo
PVInfo predicates.PersistentVolumeInfo
PVCInfo predicates.PersistentVolumeClaimInfo
@ -166,6 +167,17 @@ func InsertPredicateKeyToAlgorithmProviderMap(key string) {
return
}
// InsertPriorityKeyToAlgorithmProviderMap inserts a priority function to all algorithmProviders which are in algorithmProviderMap.
func InsertPriorityKeyToAlgorithmProviderMap(key string) {
schedulerFactoryMutex.Lock()
defer schedulerFactoryMutex.Unlock()
for _, provider := range algorithmProviderMap {
provider.PriorityFunctionKeys.Insert(key)
}
return
}
// RegisterMandatoryFitPredicate registers a fit predicate with the algorithm registry, the predicate is used by
// kubelet, DaemonSet; it is always included in configuration. Returns the name with which the predicate was
// registered.
@ -221,12 +233,12 @@ func RegisterCustomFitPredicate(policy schedulerapi.PredicatePolicy) string {
}
} else if predicateFactory, ok = fitPredicateMap[policy.Name]; ok {
// checking to see if a pre-defined predicate is requested
glog.V(2).Infof("Predicate type %s already registered, reusing.", policy.Name)
klog.V(2).Infof("Predicate type %s already registered, reusing.", policy.Name)
return policy.Name
}
if predicateFactory == nil {
glog.Fatalf("Invalid configuration: Predicate type not found for %s", policy.Name)
klog.Fatalf("Invalid configuration: Predicate type not found for %s", policy.Name)
}
return RegisterFitPredicateFactory(policy.Name, predicateFactory)
@ -333,7 +345,7 @@ func RegisterCustomPriorityFunction(policy schedulerapi.PriorityPolicy) string {
}
}
} else if existingPcf, ok := priorityFunctionMap[policy.Name]; ok {
glog.V(2).Infof("Priority type %s already registered, reusing.", policy.Name)
klog.V(2).Infof("Priority type %s already registered, reusing.", policy.Name)
// set/update the weight based on the policy
pcf = &PriorityConfigFactory{
Function: existingPcf.Function,
@ -343,7 +355,7 @@ func RegisterCustomPriorityFunction(policy schedulerapi.PriorityPolicy) string {
}
if pcf == nil {
glog.Fatalf("Invalid configuration: Priority type not found for %s", policy.Name)
klog.Fatalf("Invalid configuration: Priority type not found for %s", policy.Name)
}
return RegisterPriorityConfigFactory(policy.Name, *pcf)
@ -357,7 +369,7 @@ func buildScoringFunctionShapeFromRequestedToCapacityRatioArguments(arguments *s
}
shape, err := priorities.NewFunctionShape(points)
if err != nil {
glog.Fatalf("invalid RequestedToCapacityRatioPriority arguments: %s", err.Error())
klog.Fatalf("invalid RequestedToCapacityRatioPriority arguments: %s", err.Error())
}
return shape
}
@ -488,7 +500,7 @@ var validName = regexp.MustCompile("^[a-zA-Z0-9]([-a-zA-Z0-9]*[a-zA-Z0-9])$")
func validateAlgorithmNameOrDie(name string) {
if !validName.MatchString(name) {
glog.Fatalf("Algorithm name %v does not match the name validation regexp \"%v\".", name, validName)
klog.Fatalf("Algorithm name %v does not match the name validation regexp \"%v\".", name, validName)
}
}
@ -502,7 +514,7 @@ func validatePredicateOrDie(predicate schedulerapi.PredicatePolicy) {
numArgs++
}
if numArgs != 1 {
glog.Fatalf("Exactly 1 predicate argument is required, numArgs: %v, Predicate: %s", numArgs, predicate.Name)
klog.Fatalf("Exactly 1 predicate argument is required, numArgs: %v, Predicate: %s", numArgs, predicate.Name)
}
}
}
@ -520,7 +532,7 @@ func validatePriorityOrDie(priority schedulerapi.PriorityPolicy) {
numArgs++
}
if numArgs != 1 {
glog.Fatalf("Exactly 1 priority argument is required, numArgs: %v, Priority: %s", numArgs, priority.Name)
klog.Fatalf("Exactly 1 priority argument is required, numArgs: %v, Priority: %s", numArgs, priority.Name)
}
}
}

View File

@ -36,14 +36,18 @@ func TestAlgorithmNameValidation(t *testing.T) {
"Some,Alg:orithm",
}
for _, name := range algorithmNamesShouldValidate {
if !validName.MatchString(name) {
t.Errorf("%v should be a valid algorithm name but is not valid.", name)
}
t.Run(name, func(t *testing.T) {
if !validName.MatchString(name) {
t.Errorf("should be a valid algorithm name but is not valid.")
}
})
}
for _, name := range algorithmNamesShouldNotValidate {
if validName.MatchString(name) {
t.Errorf("%v should be an invalid algorithm name but is valid.", name)
}
t.Run(name, func(t *testing.T) {
if validName.MatchString(name) {
t.Errorf("should be an invalid algorithm name but is valid.")
}
})
}
}
@ -70,16 +74,18 @@ func TestValidatePriorityConfigOverFlow(t *testing.T) {
},
}
for _, test := range tests {
err := validateSelectedConfigs(test.configs)
if test.expected {
if err == nil {
t.Errorf("Expected Overflow for %s", test.description)
t.Run(test.description, func(t *testing.T) {
err := validateSelectedConfigs(test.configs)
if test.expected {
if err == nil {
t.Errorf("Expected Overflow")
}
} else {
if err != nil {
t.Errorf("Did not expect an overflow")
}
}
} else {
if err != nil {
t.Errorf("Did not expect an overflow for %s", test.description)
}
}
})
}
}

View File

@ -0,0 +1,64 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
go_library(
name = "go_default_library",
srcs = [
"cache.go",
"interface.go",
"node_tree.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/internal/cache",
visibility = ["//visibility:public"],
deps = [
"//pkg/features:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//pkg/util/node:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/sets:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/util/wait:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = [
"cache_test.go",
"main_test.go",
"node_tree_test.go",
],
embed = [":go_default_library"],
deps = [
"//pkg/features:go_default_library",
"//pkg/kubelet/apis:go_default_library",
"//pkg/scheduler/algorithm/priorities/util:go_default_library",
"//pkg/scheduler/cache:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/api/resource:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/types:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
"//staging/src/k8s.io/apiserver/pkg/util/feature/testing:go_default_library",
],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [
":package-srcs",
"//pkg/scheduler/internal/cache/debugger:all-srcs",
"//pkg/scheduler/internal/cache/fake:all-srcs",
],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

View File

@ -23,12 +23,13 @@ import (
"k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/labels"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/apimachinery/pkg/util/wait"
utilfeature "k8s.io/apiserver/pkg/util/feature"
"k8s.io/kubernetes/pkg/features"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
"github.com/golang/glog"
policy "k8s.io/api/policy/v1beta1"
"k8s.io/klog"
)
var (
@ -51,14 +52,16 @@ type schedulerCache struct {
period time.Duration
// This mutex guards all fields within this cache struct.
mu sync.Mutex
mu sync.RWMutex
// a set of assumed pod keys.
// The key could further be used to get an entry in podStates.
assumedPods map[string]bool
// a map from pod key to podState.
podStates map[string]*podState
nodes map[string]*NodeInfo
pdbs map[string]*policy.PodDisruptionBudget
nodes map[string]*schedulercache.NodeInfo
nodeTree *NodeTree
// A map from image name to its imageState.
imageStates map[string]*imageState
}
type podState struct {
@ -69,26 +72,42 @@ type podState struct {
bindingFinished bool
}
type imageState struct {
// Size of the image
size int64
// A set of node names for nodes having this image present
nodes sets.String
}
// createImageStateSummary returns a summarizing snapshot of the given image's state.
func (cache *schedulerCache) createImageStateSummary(state *imageState) *schedulercache.ImageStateSummary {
return &schedulercache.ImageStateSummary{
Size: state.size,
NumNodes: len(state.nodes),
}
}
func newSchedulerCache(ttl, period time.Duration, stop <-chan struct{}) *schedulerCache {
return &schedulerCache{
ttl: ttl,
period: period,
stop: stop,
nodes: make(map[string]*NodeInfo),
nodes: make(map[string]*schedulercache.NodeInfo),
nodeTree: newNodeTree(nil),
assumedPods: make(map[string]bool),
podStates: make(map[string]*podState),
pdbs: make(map[string]*policy.PodDisruptionBudget),
imageStates: make(map[string]*imageState),
}
}
// Snapshot takes a snapshot of the current schedulerCache. The method has performance impact,
// Snapshot takes a snapshot of the current schedulerinternalcache. The method has performance impact,
// and should be only used in non-critical path.
func (cache *schedulerCache) Snapshot() *Snapshot {
cache.mu.Lock()
defer cache.mu.Unlock()
cache.mu.RLock()
defer cache.mu.RUnlock()
nodes := make(map[string]*NodeInfo)
nodes := make(map[string]*schedulercache.NodeInfo)
for k, v := range cache.nodes {
nodes[k] = v.Clone()
}
@ -98,27 +117,22 @@ func (cache *schedulerCache) Snapshot() *Snapshot {
assumedPods[k] = v
}
pdbs := make(map[string]*policy.PodDisruptionBudget)
for k, v := range cache.pdbs {
pdbs[k] = v.DeepCopy()
}
return &Snapshot{
Nodes: nodes,
AssumedPods: assumedPods,
Pdbs: pdbs,
}
}
func (cache *schedulerCache) UpdateNodeNameToInfoMap(nodeNameToInfo map[string]*NodeInfo) error {
func (cache *schedulerCache) UpdateNodeNameToInfoMap(nodeNameToInfo map[string]*schedulercache.NodeInfo) error {
cache.mu.Lock()
defer cache.mu.Unlock()
for name, info := range cache.nodes {
if utilfeature.DefaultFeatureGate.Enabled(features.BalanceAttachedNodeVolumes) && info.TransientInfo != nil {
// Transient scheduler info is reset here.
info.TransientInfo.resetTransientSchedulerInfo()
info.TransientInfo.ResetTransientSchedulerInfo()
}
if current, ok := nodeNameToInfo[name]; !ok || current.generation != info.generation {
if current, ok := nodeNameToInfo[name]; !ok || current.GetGeneration() != info.GetGeneration() {
nodeNameToInfo[name] = info.Clone()
}
}
@ -136,18 +150,18 @@ func (cache *schedulerCache) List(selector labels.Selector) ([]*v1.Pod, error) {
}
func (cache *schedulerCache) FilteredList(podFilter PodFilter, selector labels.Selector) ([]*v1.Pod, error) {
cache.mu.Lock()
defer cache.mu.Unlock()
cache.mu.RLock()
defer cache.mu.RUnlock()
// podFilter is expected to return true for most or all of the pods. We
// can avoid expensive array growth without wasting too much memory by
// pre-allocating capacity.
maxSize := 0
for _, info := range cache.nodes {
maxSize += len(info.pods)
maxSize += len(info.Pods())
}
pods := make([]*v1.Pod, 0, maxSize)
for _, info := range cache.nodes {
for _, pod := range info.pods {
for _, pod := range info.Pods() {
if podFilter(pod) && selector.Matches(labels.Set(pod.Labels)) {
pods = append(pods, pod)
}
@ -157,7 +171,7 @@ func (cache *schedulerCache) FilteredList(podFilter PodFilter, selector labels.S
}
func (cache *schedulerCache) AssumePod(pod *v1.Pod) error {
key, err := getPodKey(pod)
key, err := schedulercache.GetPodKey(pod)
if err != nil {
return err
}
@ -183,15 +197,15 @@ func (cache *schedulerCache) FinishBinding(pod *v1.Pod) error {
// finishBinding exists to make tests determinitistic by injecting now as an argument
func (cache *schedulerCache) finishBinding(pod *v1.Pod, now time.Time) error {
key, err := getPodKey(pod)
key, err := schedulercache.GetPodKey(pod)
if err != nil {
return err
}
cache.mu.Lock()
defer cache.mu.Unlock()
cache.mu.RLock()
defer cache.mu.RUnlock()
glog.V(5).Infof("Finished binding for pod %v. Can be expired.", key)
klog.V(5).Infof("Finished binding for pod %v. Can be expired.", key)
currState, ok := cache.podStates[key]
if ok && cache.assumedPods[key] {
dl := now.Add(cache.ttl)
@ -202,7 +216,7 @@ func (cache *schedulerCache) finishBinding(pod *v1.Pod, now time.Time) error {
}
func (cache *schedulerCache) ForgetPod(pod *v1.Pod) error {
key, err := getPodKey(pod)
key, err := schedulercache.GetPodKey(pod)
if err != nil {
return err
}
@ -234,7 +248,7 @@ func (cache *schedulerCache) ForgetPod(pod *v1.Pod) error {
func (cache *schedulerCache) addPod(pod *v1.Pod) {
n, ok := cache.nodes[pod.Spec.NodeName]
if !ok {
n = NewNodeInfo()
n = schedulercache.NewNodeInfo()
cache.nodes[pod.Spec.NodeName] = n
}
n.AddPod(pod)
@ -255,14 +269,14 @@ func (cache *schedulerCache) removePod(pod *v1.Pod) error {
if err := n.RemovePod(pod); err != nil {
return err
}
if len(n.pods) == 0 && n.node == nil {
if len(n.Pods()) == 0 && n.Node() == nil {
delete(cache.nodes, pod.Spec.NodeName)
}
return nil
}
func (cache *schedulerCache) AddPod(pod *v1.Pod) error {
key, err := getPodKey(pod)
key, err := schedulercache.GetPodKey(pod)
if err != nil {
return err
}
@ -275,7 +289,7 @@ func (cache *schedulerCache) AddPod(pod *v1.Pod) error {
case ok && cache.assumedPods[key]:
if currState.pod.Spec.NodeName != pod.Spec.NodeName {
// The pod was added to a different node than it was assumed to.
glog.Warningf("Pod %v was assumed to be on %v but got added to %v", key, pod.Spec.NodeName, currState.pod.Spec.NodeName)
klog.Warningf("Pod %v was assumed to be on %v but got added to %v", key, pod.Spec.NodeName, currState.pod.Spec.NodeName)
// Clean this up.
cache.removePod(currState.pod)
cache.addPod(pod)
@ -297,7 +311,7 @@ func (cache *schedulerCache) AddPod(pod *v1.Pod) error {
}
func (cache *schedulerCache) UpdatePod(oldPod, newPod *v1.Pod) error {
key, err := getPodKey(oldPod)
key, err := schedulercache.GetPodKey(oldPod)
if err != nil {
return err
}
@ -311,12 +325,13 @@ func (cache *schedulerCache) UpdatePod(oldPod, newPod *v1.Pod) error {
// before Update event, in which case the state would change from Assumed to Added.
case ok && !cache.assumedPods[key]:
if currState.pod.Spec.NodeName != newPod.Spec.NodeName {
glog.Errorf("Pod %v updated on a different node than previously added to.", key)
glog.Fatalf("Schedulercache is corrupted and can badly affect scheduling decisions")
klog.Errorf("Pod %v updated on a different node than previously added to.", key)
klog.Fatalf("Schedulercache is corrupted and can badly affect scheduling decisions")
}
if err := cache.updatePod(oldPod, newPod); err != nil {
return err
}
currState.pod = newPod
default:
return fmt.Errorf("pod %v is not added to scheduler cache, so cannot be updated", key)
}
@ -324,7 +339,7 @@ func (cache *schedulerCache) UpdatePod(oldPod, newPod *v1.Pod) error {
}
func (cache *schedulerCache) RemovePod(pod *v1.Pod) error {
key, err := getPodKey(pod)
key, err := schedulercache.GetPodKey(pod)
if err != nil {
return err
}
@ -338,8 +353,8 @@ func (cache *schedulerCache) RemovePod(pod *v1.Pod) error {
// before Remove event, in which case the state would change from Assumed to Added.
case ok && !cache.assumedPods[key]:
if currState.pod.Spec.NodeName != pod.Spec.NodeName {
glog.Errorf("Pod %v was assumed to be on %v but got added to %v", key, pod.Spec.NodeName, currState.pod.Spec.NodeName)
glog.Fatalf("Schedulercache is corrupted and can badly affect scheduling decisions")
klog.Errorf("Pod %v was assumed to be on %v but got added to %v", key, pod.Spec.NodeName, currState.pod.Spec.NodeName)
klog.Fatalf("Schedulercache is corrupted and can badly affect scheduling decisions")
}
err := cache.removePod(currState.pod)
if err != nil {
@ -353,13 +368,13 @@ func (cache *schedulerCache) RemovePod(pod *v1.Pod) error {
}
func (cache *schedulerCache) IsAssumedPod(pod *v1.Pod) (bool, error) {
key, err := getPodKey(pod)
key, err := schedulercache.GetPodKey(pod)
if err != nil {
return false, err
}
cache.mu.Lock()
defer cache.mu.Unlock()
cache.mu.RLock()
defer cache.mu.RUnlock()
b, found := cache.assumedPods[key]
if !found {
@ -369,13 +384,13 @@ func (cache *schedulerCache) IsAssumedPod(pod *v1.Pod) (bool, error) {
}
func (cache *schedulerCache) GetPod(pod *v1.Pod) (*v1.Pod, error) {
key, err := getPodKey(pod)
key, err := schedulercache.GetPodKey(pod)
if err != nil {
return nil, err
}
cache.mu.Lock()
defer cache.mu.Unlock()
cache.mu.RLock()
defer cache.mu.RUnlock()
podState, ok := cache.podStates[key]
if !ok {
@ -391,9 +406,14 @@ func (cache *schedulerCache) AddNode(node *v1.Node) error {
n, ok := cache.nodes[node.Name]
if !ok {
n = NewNodeInfo()
n = schedulercache.NewNodeInfo()
cache.nodes[node.Name] = n
} else {
cache.removeNodeImageStates(n.Node())
}
cache.nodeTree.AddNode(node)
cache.addNodeImageStates(node, n)
return n.SetNode(node)
}
@ -403,9 +423,14 @@ func (cache *schedulerCache) UpdateNode(oldNode, newNode *v1.Node) error {
n, ok := cache.nodes[newNode.Name]
if !ok {
n = NewNodeInfo()
n = schedulercache.NewNodeInfo()
cache.nodes[newNode.Name] = n
} else {
cache.removeNodeImageStates(n.Node())
}
cache.nodeTree.UpdateNode(oldNode, newNode)
cache.addNodeImageStates(newNode, n)
return n.SetNode(newNode)
}
@ -421,50 +446,64 @@ func (cache *schedulerCache) RemoveNode(node *v1.Node) error {
// We can't do it unconditionally, because notifications about pods are delivered
// in a different watch, and thus can potentially be observed later, even though
// they happened before node removal.
if len(n.pods) == 0 && n.node == nil {
if len(n.Pods()) == 0 && n.Node() == nil {
delete(cache.nodes, node.Name)
}
cache.nodeTree.RemoveNode(node)
cache.removeNodeImageStates(node)
return nil
}
func (cache *schedulerCache) AddPDB(pdb *policy.PodDisruptionBudget) error {
cache.mu.Lock()
defer cache.mu.Unlock()
// addNodeImageStates adds states of the images on given node to the given nodeInfo and update the imageStates in
// scheduler cache. This function assumes the lock to scheduler cache has been acquired.
func (cache *schedulerCache) addNodeImageStates(node *v1.Node, nodeInfo *schedulercache.NodeInfo) {
newSum := make(map[string]*schedulercache.ImageStateSummary)
// Unconditionally update cache.
cache.pdbs[string(pdb.UID)] = pdb
return nil
}
func (cache *schedulerCache) UpdatePDB(oldPDB, newPDB *policy.PodDisruptionBudget) error {
return cache.AddPDB(newPDB)
}
func (cache *schedulerCache) RemovePDB(pdb *policy.PodDisruptionBudget) error {
cache.mu.Lock()
defer cache.mu.Unlock()
delete(cache.pdbs, string(pdb.UID))
return nil
}
func (cache *schedulerCache) ListPDBs(selector labels.Selector) ([]*policy.PodDisruptionBudget, error) {
cache.mu.Lock()
defer cache.mu.Unlock()
var pdbs []*policy.PodDisruptionBudget
for _, pdb := range cache.pdbs {
if selector.Matches(labels.Set(pdb.Labels)) {
pdbs = append(pdbs, pdb)
for _, image := range node.Status.Images {
for _, name := range image.Names {
// update the entry in imageStates
state, ok := cache.imageStates[name]
if !ok {
state = &imageState{
size: image.SizeBytes,
nodes: sets.NewString(node.Name),
}
cache.imageStates[name] = state
} else {
state.nodes.Insert(node.Name)
}
// create the imageStateSummary for this image
if _, ok := newSum[name]; !ok {
newSum[name] = cache.createImageStateSummary(state)
}
}
}
return pdbs, nil
nodeInfo.SetImageStates(newSum)
}
func (cache *schedulerCache) IsUpToDate(n *NodeInfo) bool {
cache.mu.Lock()
defer cache.mu.Unlock()
node, ok := cache.nodes[n.Node().Name]
return ok && n.generation == node.generation
// removeNodeImageStates removes the given node record from image entries having the node
// in imageStates cache. After the removal, if any image becomes free, i.e., the image
// is no longer available on any node, the image entry will be removed from imageStates.
func (cache *schedulerCache) removeNodeImageStates(node *v1.Node) {
if node == nil {
return
}
for _, image := range node.Status.Images {
for _, name := range image.Names {
state, ok := cache.imageStates[name]
if ok {
state.nodes.Delete(node.Name)
if len(state.nodes) == 0 {
// Remove the unused image to make sure the length of
// imageStates represents the total number of different
// images on all nodes
delete(cache.imageStates, name)
}
}
}
}
}
func (cache *schedulerCache) run() {
@ -487,14 +526,14 @@ func (cache *schedulerCache) cleanupAssumedPods(now time.Time) {
panic("Key found in assumed set but not in podStates. Potentially a logical error.")
}
if !ps.bindingFinished {
glog.V(3).Infof("Couldn't expire cache for pod %v/%v. Binding is still in progress.",
klog.V(3).Infof("Couldn't expire cache for pod %v/%v. Binding is still in progress.",
ps.pod.Namespace, ps.pod.Name)
continue
}
if now.After(*ps.deadline) {
glog.Warningf("Pod %s/%s expired", ps.pod.Namespace, ps.pod.Name)
klog.Warningf("Pod %s/%s expired", ps.pod.Namespace, ps.pod.Name)
if err := cache.expirePod(key, ps); err != nil {
glog.Errorf("ExpirePod failed for %s: %v", key, err)
klog.Errorf("ExpirePod failed for %s: %v", key, err)
}
}
}
@ -508,3 +547,7 @@ func (cache *schedulerCache) expirePod(key string, ps *podState) error {
delete(cache.podStates, key)
return nil
}
func (cache *schedulerCache) NodeTree() *NodeTree {
return cache.nodeTree
}

View File

@ -24,23 +24,24 @@ import (
"time"
"k8s.io/api/core/v1"
"k8s.io/api/policy/v1beta1"
"k8s.io/apimachinery/pkg/api/resource"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/labels"
"k8s.io/apimachinery/pkg/types"
"k8s.io/apimachinery/pkg/util/intstr"
"k8s.io/apimachinery/pkg/util/wait"
utilfeature "k8s.io/apiserver/pkg/util/feature"
utilfeaturetesting "k8s.io/apiserver/pkg/util/feature/testing"
"k8s.io/kubernetes/pkg/features"
priorityutil "k8s.io/kubernetes/pkg/scheduler/algorithm/priorities/util"
schedutil "k8s.io/kubernetes/pkg/scheduler/util"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
)
func deepEqualWithoutGeneration(t *testing.T, testcase int, actual, expected *NodeInfo) {
func deepEqualWithoutGeneration(t *testing.T, testcase int, actual, expected *schedulercache.NodeInfo) {
// Ignore generation field.
if actual != nil {
actual.generation = 0
actual.SetGeneration(0)
}
if expected != nil {
expected.SetGeneration(0)
}
if !reflect.DeepEqual(actual, expected) {
t.Errorf("#%d: node info get=%s, want=%s", testcase, actual, expected)
@ -65,19 +66,34 @@ func (b *hostPortInfoBuilder) add(protocol, ip string, port int32) *hostPortInfo
return b
}
func (b *hostPortInfoBuilder) build() schedutil.HostPortInfo {
res := make(schedutil.HostPortInfo)
func (b *hostPortInfoBuilder) build() schedulercache.HostPortInfo {
res := make(schedulercache.HostPortInfo)
for _, param := range b.inputs {
res.Add(param.ip, param.protocol, param.port)
}
return res
}
func newNodeInfo(requestedResource *schedulercache.Resource,
nonzeroRequest *schedulercache.Resource,
pods []*v1.Pod,
usedPorts schedulercache.HostPortInfo,
imageStates map[string]*schedulercache.ImageStateSummary,
) *schedulercache.NodeInfo {
nodeInfo := schedulercache.NewNodeInfo(pods...)
nodeInfo.SetRequestedResource(requestedResource)
nodeInfo.SetNonZeroRequest(nonzeroRequest)
nodeInfo.SetUsedPorts(usedPorts)
nodeInfo.SetImageStates(imageStates)
return nodeInfo
}
// TestAssumePodScheduled tests that after a pod is assumed, its information is aggregated
// on node level.
func TestAssumePodScheduled(t *testing.T) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
nodeName := "node"
testPods := []*v1.Pod{
makeBasePod(t, nodeName, "test", "100m", "500", "", []v1.ContainerPort{{HostIP: "127.0.0.1", HostPort: 80, Protocol: "TCP"}}),
@ -92,111 +108,99 @@ func TestAssumePodScheduled(t *testing.T) {
tests := []struct {
pods []*v1.Pod
wNodeInfo *NodeInfo
wNodeInfo *schedulercache.NodeInfo
}{{
pods: []*v1.Pod{testPods[0]},
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[0]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{testPods[0]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
}, {
pods: []*v1.Pod{testPods[1], testPods[2]},
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 300,
Memory: 1524,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 300,
Memory: 1524,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[1], testPods[2]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).add("TCP", "127.0.0.1", 8080).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{testPods[1], testPods[2]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).add("TCP", "127.0.0.1", 8080).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
}, { // test non-zero request
pods: []*v1.Pod{testPods[3]},
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 0,
Memory: 0,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: priorityutil.DefaultMilliCPURequest,
Memory: priorityutil.DefaultMemoryRequest,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[3]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{testPods[3]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
}, {
pods: []*v1.Pod{testPods[4]},
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
ScalarResources: map[v1.ResourceName]int64{"example.com/foo": 3},
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[4]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{testPods[4]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
}, {
pods: []*v1.Pod{testPods[4], testPods[5]},
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 300,
Memory: 1524,
ScalarResources: map[v1.ResourceName]int64{"example.com/foo": 8},
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 300,
Memory: 1524,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[4], testPods[5]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).add("TCP", "127.0.0.1", 8080).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{testPods[4], testPods[5]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).add("TCP", "127.0.0.1", 8080).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
}, {
pods: []*v1.Pod{testPods[6]},
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[6]},
usedPorts: newHostPortInfoBuilder().build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{testPods[6]},
newHostPortInfoBuilder().build(),
make(map[string]*schedulercache.ImageStateSummary),
),
},
}
@ -237,7 +241,7 @@ func assumeAndFinishBinding(cache *schedulerCache, pod *v1.Pod, assumedTime time
// The removal will be reflected in node info.
func TestExpirePod(t *testing.T) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
nodeName := "node"
testPods := []*v1.Pod{
makeBasePod(t, nodeName, "test-1", "100m", "500", "", []v1.ContainerPort{{HostIP: "127.0.0.1", HostPort: 80, Protocol: "TCP"}}),
@ -249,7 +253,7 @@ func TestExpirePod(t *testing.T) {
pods []*testExpirePodStruct
cleanupTime time.Time
wNodeInfo *NodeInfo
wNodeInfo *schedulercache.NodeInfo
}{{ // assumed pod would expires
pods: []*testExpirePodStruct{
{pod: testPods[0], assumedTime: now},
@ -262,21 +266,19 @@ func TestExpirePod(t *testing.T) {
{pod: testPods[1], assumedTime: now.Add(3 * ttl / 2)},
},
cleanupTime: now.Add(2 * ttl),
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 200,
Memory: 1024,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 200,
Memory: 1024,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[1]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 8080).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{testPods[1]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 8080).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
}}
for i, tt := range tests {
@ -298,7 +300,7 @@ func TestExpirePod(t *testing.T) {
// The pod info should still exist after manually expiring unconfirmed pods.
func TestAddPodWillConfirm(t *testing.T) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
nodeName := "node"
now := time.Now()
ttl := 10 * time.Second
@ -311,25 +313,23 @@ func TestAddPodWillConfirm(t *testing.T) {
podsToAssume []*v1.Pod
podsToAdd []*v1.Pod
wNodeInfo *NodeInfo
wNodeInfo *schedulercache.NodeInfo
}{{ // two pod were assumed at same time. But first one is called Add() and gets confirmed.
podsToAssume: []*v1.Pod{testPods[0], testPods[1]},
podsToAdd: []*v1.Pod{testPods[0]},
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[0]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{testPods[0]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
}}
for i, tt := range tests {
@ -405,28 +405,26 @@ func TestAddPodWillReplaceAssumed(t *testing.T) {
podsToAdd []*v1.Pod
podsToUpdate [][]*v1.Pod
wNodeInfo map[string]*NodeInfo
wNodeInfo map[string]*schedulercache.NodeInfo
}{{
podsToAssume: []*v1.Pod{assumedPod.DeepCopy()},
podsToAdd: []*v1.Pod{addedPod.DeepCopy()},
podsToUpdate: [][]*v1.Pod{{addedPod.DeepCopy(), updatedPod.DeepCopy()}},
wNodeInfo: map[string]*NodeInfo{
wNodeInfo: map[string]*schedulercache.NodeInfo{
"assumed-node": nil,
"actual-node": {
requestedResource: &Resource{
"actual-node": newNodeInfo(
&schedulercache.Resource{
MilliCPU: 200,
Memory: 500,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 200,
Memory: 500,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{updatedPod.DeepCopy()},
usedPorts: newHostPortInfoBuilder().add("TCP", "0.0.0.0", 90).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{updatedPod.DeepCopy()},
newHostPortInfoBuilder().add("TCP", "0.0.0.0", 90).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
},
}}
@ -458,31 +456,29 @@ func TestAddPodWillReplaceAssumed(t *testing.T) {
// TestAddPodAfterExpiration tests that a pod being Add()ed will be added back if expired.
func TestAddPodAfterExpiration(t *testing.T) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
nodeName := "node"
ttl := 10 * time.Second
basePod := makeBasePod(t, nodeName, "test", "100m", "500", "", []v1.ContainerPort{{HostIP: "127.0.0.1", HostPort: 80, Protocol: "TCP"}})
tests := []struct {
pod *v1.Pod
wNodeInfo *NodeInfo
wNodeInfo *schedulercache.NodeInfo
}{{
pod: basePod,
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{basePod},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{basePod},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
}}
now := time.Now()
@ -509,7 +505,7 @@ func TestAddPodAfterExpiration(t *testing.T) {
// TestUpdatePod tests that a pod will be updated if added before.
func TestUpdatePod(t *testing.T) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
nodeName := "node"
ttl := 10 * time.Second
testPods := []*v1.Pod{
@ -520,39 +516,35 @@ func TestUpdatePod(t *testing.T) {
podsToAdd []*v1.Pod
podsToUpdate []*v1.Pod
wNodeInfo []*NodeInfo
wNodeInfo []*schedulercache.NodeInfo
}{{ // add a pod and then update it twice
podsToAdd: []*v1.Pod{testPods[0]},
podsToUpdate: []*v1.Pod{testPods[0], testPods[1], testPods[0]},
wNodeInfo: []*NodeInfo{{
requestedResource: &Resource{
wNodeInfo: []*schedulercache.NodeInfo{newNodeInfo(
&schedulercache.Resource{
MilliCPU: 200,
Memory: 1024,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 200,
Memory: 1024,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[1]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 8080).build(),
imageSizes: map[string]int64{},
}, {
requestedResource: &Resource{
[]*v1.Pod{testPods[1]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 8080).build(),
make(map[string]*schedulercache.ImageStateSummary),
), newNodeInfo(
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[0]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
imageSizes: map[string]int64{},
}},
[]*v1.Pod{testPods[0]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
make(map[string]*schedulercache.ImageStateSummary),
)},
}}
for _, tt := range tests {
@ -577,10 +569,69 @@ func TestUpdatePod(t *testing.T) {
}
}
// TestUpdatePodAndGet tests get always return latest pod state
func TestUpdatePodAndGet(t *testing.T) {
nodeName := "node"
ttl := 10 * time.Second
testPods := []*v1.Pod{
makeBasePod(t, nodeName, "test", "100m", "500", "", []v1.ContainerPort{{HostIP: "127.0.0.1", HostPort: 80, Protocol: "TCP"}}),
makeBasePod(t, nodeName, "test", "200m", "1Ki", "", []v1.ContainerPort{{HostIP: "127.0.0.1", HostPort: 8080, Protocol: "TCP"}}),
}
tests := []struct {
pod *v1.Pod
podToUpdate *v1.Pod
handler func(cache Cache, pod *v1.Pod) error
assumePod bool
}{
{
pod: testPods[0],
podToUpdate: testPods[0],
handler: func(cache Cache, pod *v1.Pod) error {
return cache.AssumePod(pod)
},
assumePod: true,
},
{
pod: testPods[0],
podToUpdate: testPods[1],
handler: func(cache Cache, pod *v1.Pod) error {
return cache.AddPod(pod)
},
assumePod: false,
},
}
for _, tt := range tests {
cache := newSchedulerCache(ttl, time.Second, nil)
if err := tt.handler(cache, tt.pod); err != nil {
t.Fatalf("unexpected err: %v", err)
}
if !tt.assumePod {
if err := cache.UpdatePod(tt.pod, tt.podToUpdate); err != nil {
t.Fatalf("UpdatePod failed: %v", err)
}
}
cachedPod, err := cache.GetPod(tt.pod)
if err != nil {
t.Fatalf("GetPod failed: %v", err)
}
if !reflect.DeepEqual(tt.podToUpdate, cachedPod) {
t.Fatalf("pod get=%s, want=%s", cachedPod, tt.podToUpdate)
}
}
}
// TestExpireAddUpdatePod test the sequence that a pod is expired, added, then updated
func TestExpireAddUpdatePod(t *testing.T) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
nodeName := "node"
ttl := 10 * time.Second
testPods := []*v1.Pod{
@ -592,40 +643,36 @@ func TestExpireAddUpdatePod(t *testing.T) {
podsToAdd []*v1.Pod
podsToUpdate []*v1.Pod
wNodeInfo []*NodeInfo
wNodeInfo []*schedulercache.NodeInfo
}{{ // Pod is assumed, expired, and added. Then it would be updated twice.
podsToAssume: []*v1.Pod{testPods[0]},
podsToAdd: []*v1.Pod{testPods[0]},
podsToUpdate: []*v1.Pod{testPods[0], testPods[1], testPods[0]},
wNodeInfo: []*NodeInfo{{
requestedResource: &Resource{
wNodeInfo: []*schedulercache.NodeInfo{newNodeInfo(
&schedulercache.Resource{
MilliCPU: 200,
Memory: 1024,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 200,
Memory: 1024,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[1]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 8080).build(),
imageSizes: map[string]int64{},
}, {
requestedResource: &Resource{
[]*v1.Pod{testPods[1]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 8080).build(),
make(map[string]*schedulercache.ImageStateSummary),
), newNodeInfo(
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{testPods[0]},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
imageSizes: map[string]int64{},
}},
[]*v1.Pod{testPods[0]},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
make(map[string]*schedulercache.ImageStateSummary),
)},
}}
now := time.Now()
@ -681,29 +728,27 @@ func makePodWithEphemeralStorage(nodeName, ephemeralStorage string) *v1.Pod {
func TestEphemeralStorageResource(t *testing.T) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
nodeName := "node"
podE := makePodWithEphemeralStorage(nodeName, "500")
tests := []struct {
pod *v1.Pod
wNodeInfo *NodeInfo
wNodeInfo *schedulercache.NodeInfo
}{
{
pod: podE,
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
EphemeralStorage: 500,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: priorityutil.DefaultMilliCPURequest,
Memory: priorityutil.DefaultMemoryRequest,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{podE},
usedPorts: schedutil.HostPortInfo{},
imageSizes: map[string]int64{},
},
[]*v1.Pod{podE},
schedulercache.HostPortInfo{},
make(map[string]*schedulercache.ImageStateSummary),
),
},
}
for i, tt := range tests {
@ -728,29 +773,27 @@ func TestEphemeralStorageResource(t *testing.T) {
// TestRemovePod tests after added pod is removed, its information should also be subtracted.
func TestRemovePod(t *testing.T) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(t, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
nodeName := "node"
basePod := makeBasePod(t, nodeName, "test", "100m", "500", "", []v1.ContainerPort{{HostIP: "127.0.0.1", HostPort: 80, Protocol: "TCP"}})
tests := []struct {
pod *v1.Pod
wNodeInfo *NodeInfo
wNodeInfo *schedulercache.NodeInfo
}{{
pod: basePod,
wNodeInfo: &NodeInfo{
requestedResource: &Resource{
wNodeInfo: newNodeInfo(
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
nonzeroRequest: &Resource{
&schedulercache.Resource{
MilliCPU: 100,
Memory: 500,
},
TransientInfo: newTransientSchedulerInfo(),
allocatableResource: &Resource{},
pods: []*v1.Pod{basePod},
usedPorts: newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
imageSizes: map[string]int64{},
},
[]*v1.Pod{basePod},
newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
make(map[string]*schedulercache.ImageStateSummary),
),
}}
for i, tt := range tests {
@ -829,7 +872,7 @@ func TestForgetPod(t *testing.T) {
// getResourceRequest returns the resource request of all containers in Pods;
// excuding initContainers.
func getResourceRequest(pod *v1.Pod) v1.ResourceList {
result := &Resource{}
result := &schedulercache.Resource{}
for _, container := range pod.Spec.Containers {
result.Add(container.Resources.Requests)
}
@ -838,22 +881,30 @@ func getResourceRequest(pod *v1.Pod) v1.ResourceList {
}
// buildNodeInfo creates a NodeInfo by simulating node operations in cache.
func buildNodeInfo(node *v1.Node, pods []*v1.Pod) *NodeInfo {
expected := NewNodeInfo()
func buildNodeInfo(node *v1.Node, pods []*v1.Pod) *schedulercache.NodeInfo {
expected := schedulercache.NewNodeInfo()
// Simulate SetNode.
expected.node = node
expected.allocatableResource = NewResource(node.Status.Allocatable)
expected.taints = node.Spec.Taints
expected.generation++
expected.SetNode(node)
expected.SetAllocatableResource(schedulercache.NewResource(node.Status.Allocatable))
expected.SetTaints(node.Spec.Taints)
expected.SetGeneration(expected.GetGeneration() + 1)
for _, pod := range pods {
// Simulate AddPod
expected.pods = append(expected.pods, pod)
expected.requestedResource.Add(getResourceRequest(pod))
expected.nonzeroRequest.Add(getResourceRequest(pod))
expected.updateUsedPorts(pod, true)
expected.generation++
pods := append(expected.Pods(), pod)
expected.SetPods(pods)
requestedResource := expected.RequestedResource()
newRequestedResource := &requestedResource
newRequestedResource.Add(getResourceRequest(pod))
expected.SetRequestedResource(newRequestedResource)
nonZeroRequest := expected.NonZeroRequest()
newNonZeroRequest := &nonZeroRequest
newNonZeroRequest.Add(getResourceRequest(pod))
expected.SetNonZeroRequest(newNonZeroRequest)
expected.UpdateUsedPorts(pod, true)
expected.SetGeneration(expected.GetGeneration() + 1)
}
return expected
@ -1004,49 +1055,65 @@ func TestNodeOperators(t *testing.T) {
// Case 1: the node was added into cache successfully.
got, found := cache.nodes[node.Name]
if !found {
t.Errorf("Failed to find node %v in schedulercache.", node.Name)
t.Errorf("Failed to find node %v in schedulerinternalcache.", node.Name)
}
if cache.nodeTree.NumNodes != 1 || cache.nodeTree.Next() != node.Name {
t.Errorf("cache.nodeTree is not updated correctly after adding node: %v", node.Name)
}
// Generations are globally unique. We check in our unit tests that they are incremented correctly.
expected.generation = got.generation
expected.SetGeneration(got.GetGeneration())
if !reflect.DeepEqual(got, expected) {
t.Errorf("Failed to add node into schedulercache:\n got: %+v \nexpected: %+v", got, expected)
}
// Case 2: dump cached nodes successfully.
cachedNodes := map[string]*NodeInfo{}
cachedNodes := map[string]*schedulercache.NodeInfo{}
cache.UpdateNodeNameToInfoMap(cachedNodes)
newNode, found := cachedNodes[node.Name]
if !found || len(cachedNodes) != 1 {
t.Errorf("failed to dump cached nodes:\n got: %v \nexpected: %v", cachedNodes, cache.nodes)
}
expected.generation = newNode.generation
expected.SetGeneration(newNode.GetGeneration())
if !reflect.DeepEqual(newNode, expected) {
t.Errorf("Failed to clone node:\n got: %+v, \n expected: %+v", newNode, expected)
}
// Case 3: update node attribute successfully.
node.Status.Allocatable[v1.ResourceMemory] = mem50m
expected.allocatableResource.Memory = mem50m.Value()
allocatableResource := expected.AllocatableResource()
newAllocatableResource := &allocatableResource
newAllocatableResource.Memory = mem50m.Value()
expected.SetAllocatableResource(newAllocatableResource)
cache.UpdateNode(nil, node)
got, found = cache.nodes[node.Name]
if !found {
t.Errorf("Failed to find node %v in schedulercache after UpdateNode.", node.Name)
}
if got.generation <= expected.generation {
t.Errorf("generation is not incremented. got: %v, expected: %v", got.generation, expected.generation)
if got.GetGeneration() <= expected.GetGeneration() {
t.Errorf("Generation is not incremented. got: %v, expected: %v", got.GetGeneration(), expected.GetGeneration())
}
expected.generation = got.generation
expected.SetGeneration(got.GetGeneration())
if !reflect.DeepEqual(got, expected) {
t.Errorf("Failed to update node in schedulercache:\n got: %+v \nexpected: %+v", got, expected)
}
// Check nodeTree after update
if cache.nodeTree.NumNodes != 1 || cache.nodeTree.Next() != node.Name {
t.Errorf("unexpected cache.nodeTree after updating node: %v", node.Name)
}
// Case 4: the node can not be removed if pods is not empty.
cache.RemoveNode(node)
if _, found := cache.nodes[node.Name]; !found {
t.Errorf("The node %v should not be removed if pods is not empty.", node.Name)
}
// Check nodeTree after remove. The node should be removed from the nodeTree even if there are
// still pods on it.
if cache.nodeTree.NumNodes != 0 || cache.nodeTree.Next() != "" {
t.Errorf("unexpected cache.nodeTree after removing node: %v", node.Name)
}
}
}
@ -1060,25 +1127,27 @@ func BenchmarkList1kNodes30kPods(b *testing.B) {
func BenchmarkUpdate1kNodes30kPods(b *testing.B) {
// Enable volumesOnNodeForBalancing to do balanced resource allocation
utilfeature.DefaultFeatureGate.Set(fmt.Sprintf("%s=true", features.BalanceAttachedNodeVolumes))
defer utilfeaturetesting.SetFeatureGateDuringTest(nil, utilfeature.DefaultFeatureGate, features.BalanceAttachedNodeVolumes, true)()
cache := setupCacheOf1kNodes30kPods(b)
b.ResetTimer()
for n := 0; n < b.N; n++ {
cachedNodes := map[string]*NodeInfo{}
cachedNodes := map[string]*schedulercache.NodeInfo{}
cache.UpdateNodeNameToInfoMap(cachedNodes)
}
}
func BenchmarkExpire100Pods(b *testing.B) {
benchmarkExpire(b, 100)
}
func BenchmarkExpire1kPods(b *testing.B) {
benchmarkExpire(b, 1000)
}
func BenchmarkExpire10kPods(b *testing.B) {
benchmarkExpire(b, 10000)
func BenchmarkExpirePods(b *testing.B) {
podNums := []int{
100,
1000,
10000,
}
for _, podNum := range podNums {
name := fmt.Sprintf("%dPods", podNum)
b.Run(name, func(b *testing.B) {
benchmarkExpire(b, podNum)
})
}
}
func benchmarkExpire(b *testing.B, podNum int) {
@ -1158,131 +1227,3 @@ func setupCacheWithAssumedPods(b *testing.B, podNum int, assumedTime time.Time)
}
return cache
}
func makePDB(name, namespace string, uid types.UID, labels map[string]string, minAvailable int) *v1beta1.PodDisruptionBudget {
intstrMin := intstr.FromInt(minAvailable)
pdb := &v1beta1.PodDisruptionBudget{
ObjectMeta: metav1.ObjectMeta{
Namespace: namespace,
Name: name,
Labels: labels,
UID: uid,
},
Spec: v1beta1.PodDisruptionBudgetSpec{
MinAvailable: &intstrMin,
Selector: &metav1.LabelSelector{MatchLabels: labels},
},
}
return pdb
}
// TestPDBOperations tests that a PDB will be add/updated/deleted correctly.
func TestPDBOperations(t *testing.T) {
ttl := 10 * time.Second
testPDBs := []*v1beta1.PodDisruptionBudget{
makePDB("pdb0", "ns1", "uid0", map[string]string{"tkey1": "tval1"}, 3),
makePDB("pdb1", "ns1", "uid1", map[string]string{"tkey1": "tval1", "tkey2": "tval2"}, 1),
makePDB("pdb2", "ns3", "uid2", map[string]string{"tkey3": "tval3", "tkey2": "tval2"}, 10),
}
updatedPDBs := []*v1beta1.PodDisruptionBudget{
makePDB("pdb0", "ns1", "uid0", map[string]string{"tkey4": "tval4"}, 8),
makePDB("pdb1", "ns1", "uid1", map[string]string{"tkey1": "tval1"}, 1),
makePDB("pdb2", "ns3", "uid2", map[string]string{"tkey3": "tval3", "tkey1": "tval1", "tkey2": "tval2"}, 10),
}
tests := []struct {
pdbsToAdd []*v1beta1.PodDisruptionBudget
pdbsToUpdate []*v1beta1.PodDisruptionBudget
pdbsToDelete []*v1beta1.PodDisruptionBudget
expectedPDBs []*v1beta1.PodDisruptionBudget // Expected PDBs after all operations
}{
{
pdbsToAdd: []*v1beta1.PodDisruptionBudget{testPDBs[0]},
pdbsToUpdate: []*v1beta1.PodDisruptionBudget{testPDBs[0], testPDBs[1], testPDBs[0]},
expectedPDBs: []*v1beta1.PodDisruptionBudget{testPDBs[0], testPDBs[1]}, // both will be in the cache as they have different names
},
{
pdbsToAdd: []*v1beta1.PodDisruptionBudget{testPDBs[0]},
pdbsToUpdate: []*v1beta1.PodDisruptionBudget{testPDBs[0], updatedPDBs[0]},
expectedPDBs: []*v1beta1.PodDisruptionBudget{updatedPDBs[0]},
},
{
pdbsToAdd: []*v1beta1.PodDisruptionBudget{testPDBs[0], testPDBs[2]},
pdbsToUpdate: []*v1beta1.PodDisruptionBudget{testPDBs[0], updatedPDBs[0]},
pdbsToDelete: []*v1beta1.PodDisruptionBudget{testPDBs[0]},
expectedPDBs: []*v1beta1.PodDisruptionBudget{testPDBs[2]},
},
}
for _, test := range tests {
cache := newSchedulerCache(ttl, time.Second, nil)
for _, pdbToAdd := range test.pdbsToAdd {
if err := cache.AddPDB(pdbToAdd); err != nil {
t.Fatalf("AddPDB failed: %v", err)
}
}
for i := range test.pdbsToUpdate {
if i == 0 {
continue
}
if err := cache.UpdatePDB(test.pdbsToUpdate[i-1], test.pdbsToUpdate[i]); err != nil {
t.Fatalf("UpdatePDB failed: %v", err)
}
}
for _, pdb := range test.pdbsToDelete {
if err := cache.RemovePDB(pdb); err != nil {
t.Fatalf("RemovePDB failed: %v", err)
}
}
cachedPDBs, err := cache.ListPDBs(labels.Everything())
if err != nil {
t.Fatalf("ListPDBs failed: %v", err)
}
if len(cachedPDBs) != len(test.expectedPDBs) {
t.Errorf("Expected %d PDBs, got %d", len(test.expectedPDBs), len(cachedPDBs))
}
for _, pdb := range test.expectedPDBs {
found := false
// find it among the cached ones
for _, cpdb := range cachedPDBs {
if pdb.UID == cpdb.UID {
found = true
if !reflect.DeepEqual(pdb, cpdb) {
t.Errorf("%v is not equal to %v", pdb, cpdb)
}
break
}
}
if !found {
t.Errorf("PDB with uid '%v' was not found in the cache.", pdb.UID)
}
}
}
}
func TestIsUpToDate(t *testing.T) {
cache := New(time.Duration(0), wait.NeverStop)
if err := cache.AddNode(&v1.Node{ObjectMeta: metav1.ObjectMeta{Name: "n1"}}); err != nil {
t.Errorf("Could not add node: %v", err)
}
s := cache.Snapshot()
node := s.Nodes["n1"]
if !cache.IsUpToDate(node) {
t.Errorf("Node incorrectly marked as stale")
}
pod := &v1.Pod{ObjectMeta: metav1.ObjectMeta{Name: "p1", UID: "p1"}, Spec: v1.PodSpec{NodeName: "n1"}}
if err := cache.AddPod(pod); err != nil {
t.Errorf("Could not add pod: %v", err)
}
if cache.IsUpToDate(node) {
t.Errorf("Node incorrectly marked as up to date")
}
badNode := &NodeInfo{node: &v1.Node{ObjectMeta: metav1.ObjectMeta{Name: "n2"}}}
if cache.IsUpToDate(badNode) {
t.Errorf("Nonexistant node incorrectly marked as up to date")
}
}

View File

@ -0,0 +1,46 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
go_library(
name = "go_default_library",
srcs = [
"comparer.go",
"debugger.go",
"dumper.go",
],
importpath = "k8s.io/kubernetes/pkg/scheduler/internal/cache/debugger",
visibility = ["//pkg/scheduler:__subpackages__"],
deps = [
"//pkg/scheduler/cache:go_default_library",
"//pkg/scheduler/internal/cache:go_default_library",
"//pkg/scheduler/internal/queue:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
"//staging/src/k8s.io/client-go/listers/core/v1:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
go_test(
name = "go_default_test",
srcs = ["comparer_test.go"],
embed = [":go_default_library"],
deps = [
"//pkg/scheduler/cache:go_default_library",
"//staging/src/k8s.io/api/core/v1:go_default_library",
"//staging/src/k8s.io/apimachinery/pkg/types:go_default_library",
],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [":package-srcs"],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

View File

@ -14,74 +14,61 @@ See the License for the specific language governing permissions and
limitations under the License.
*/
package factory
package debugger
import (
"sort"
"strings"
"github.com/golang/glog"
"k8s.io/api/core/v1"
policy "k8s.io/api/policy/v1beta1"
"k8s.io/apimachinery/pkg/labels"
corelisters "k8s.io/client-go/listers/core/v1"
v1beta1 "k8s.io/client-go/listers/policy/v1beta1"
"k8s.io/klog"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
"k8s.io/kubernetes/pkg/scheduler/core"
schedulerinternalcache "k8s.io/kubernetes/pkg/scheduler/internal/cache"
internalqueue "k8s.io/kubernetes/pkg/scheduler/internal/queue"
)
type cacheComparer struct {
nodeLister corelisters.NodeLister
podLister corelisters.PodLister
pdbLister v1beta1.PodDisruptionBudgetLister
cache schedulercache.Cache
podQueue core.SchedulingQueue
compareStrategy
// CacheComparer is an implementation of the Scheduler's cache comparer.
type CacheComparer struct {
NodeLister corelisters.NodeLister
PodLister corelisters.PodLister
Cache schedulerinternalcache.Cache
PodQueue internalqueue.SchedulingQueue
}
func (c *cacheComparer) Compare() error {
glog.V(3).Info("cache comparer started")
defer glog.V(3).Info("cache comparer finished")
// Compare compares the nodes and pods of NodeLister with Cache.Snapshot.
func (c *CacheComparer) Compare() error {
klog.V(3).Info("cache comparer started")
defer klog.V(3).Info("cache comparer finished")
nodes, err := c.nodeLister.List(labels.Everything())
nodes, err := c.NodeLister.List(labels.Everything())
if err != nil {
return err
}
pods, err := c.podLister.List(labels.Everything())
pods, err := c.PodLister.List(labels.Everything())
if err != nil {
return err
}
pdbs, err := c.pdbLister.List(labels.Everything())
if err != nil {
return err
}
snapshot := c.Cache.Snapshot()
snapshot := c.cache.Snapshot()
waitingPods := c.podQueue.WaitingPods()
waitingPods := c.PodQueue.WaitingPods()
if missed, redundant := c.CompareNodes(nodes, snapshot.Nodes); len(missed)+len(redundant) != 0 {
glog.Warningf("cache mismatch: missed nodes: %s; redundant nodes: %s", missed, redundant)
klog.Warningf("cache mismatch: missed nodes: %s; redundant nodes: %s", missed, redundant)
}
if missed, redundant := c.ComparePods(pods, waitingPods, snapshot.Nodes); len(missed)+len(redundant) != 0 {
glog.Warningf("cache mismatch: missed pods: %s; redundant pods: %s", missed, redundant)
}
if missed, redundant := c.ComparePdbs(pdbs, snapshot.Pdbs); len(missed)+len(redundant) != 0 {
glog.Warningf("cache mismatch: missed pdbs: %s; redundant pdbs: %s", missed, redundant)
klog.Warningf("cache mismatch: missed pods: %s; redundant pods: %s", missed, redundant)
}
return nil
}
type compareStrategy struct {
}
func (c compareStrategy) CompareNodes(nodes []*v1.Node, nodeinfos map[string]*schedulercache.NodeInfo) (missed, redundant []string) {
// CompareNodes compares actual nodes with cached nodes.
func (c *CacheComparer) CompareNodes(nodes []*v1.Node, nodeinfos map[string]*schedulercache.NodeInfo) (missed, redundant []string) {
actual := []string{}
for _, node := range nodes {
actual = append(actual, node.Name)
@ -95,7 +82,8 @@ func (c compareStrategy) CompareNodes(nodes []*v1.Node, nodeinfos map[string]*sc
return compareStrings(actual, cached)
}
func (c compareStrategy) ComparePods(pods, waitingPods []*v1.Pod, nodeinfos map[string]*schedulercache.NodeInfo) (missed, redundant []string) {
// ComparePods compares actual pods with cached pods.
func (c *CacheComparer) ComparePods(pods, waitingPods []*v1.Pod, nodeinfos map[string]*schedulercache.NodeInfo) (missed, redundant []string) {
actual := []string{}
for _, pod := range pods {
actual = append(actual, string(pod.UID))
@ -114,20 +102,6 @@ func (c compareStrategy) ComparePods(pods, waitingPods []*v1.Pod, nodeinfos map[
return compareStrings(actual, cached)
}
func (c compareStrategy) ComparePdbs(pdbs []*policy.PodDisruptionBudget, pdbCache map[string]*policy.PodDisruptionBudget) (missed, redundant []string) {
actual := []string{}
for _, pdb := range pdbs {
actual = append(actual, string(pdb.UID))
}
cached := []string{}
for pdbUID := range pdbCache {
cached = append(cached, pdbUID)
}
return compareStrings(actual, cached)
}
func compareStrings(actual, cached []string) (missed, redundant []string) {
missed, redundant = []string{}, []string{}

View File

@ -0,0 +1,192 @@
/*
Copyright 2018 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 debugger
import (
"reflect"
"testing"
"k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/types"
schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
)
func TestCompareNodes(t *testing.T) {
tests := []struct {
name string
actual []string
cached []string
missing []string
redundant []string
}{
{
name: "redundant cached value",
actual: []string{"foo", "bar"},
cached: []string{"bar", "foo", "foobar"},
missing: []string{},
redundant: []string{"foobar"},
},
{
name: "missing cached value",
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foo"},
missing: []string{"foobar"},
redundant: []string{},
},
{
name: "proper cache set",
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foobar", "foo"},
missing: []string{},
redundant: []string{},
},
}
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
testCompareNodes(test.actual, test.cached, test.missing, test.redundant, t)
})
}
}
func testCompareNodes(actual, cached, missing, redundant []string, t *testing.T) {
compare := CacheComparer{}
nodes := []*v1.Node{}
for _, nodeName := range actual {
node := &v1.Node{}
node.Name = nodeName
nodes = append(nodes, node)
}
nodeInfo := make(map[string]*schedulercache.NodeInfo)
for _, nodeName := range cached {
nodeInfo[nodeName] = &schedulercache.NodeInfo{}
}
m, r := compare.CompareNodes(nodes, nodeInfo)
if !reflect.DeepEqual(m, missing) {
t.Errorf("missing expected to be %s; got %s", missing, m)
}
if !reflect.DeepEqual(r, redundant) {
t.Errorf("redundant expected to be %s; got %s", redundant, r)
}
}
func TestComparePods(t *testing.T) {
tests := []struct {
name string
actual []string
cached []string
queued []string
missing []string
redundant []string
}{
{
name: "redundant cached value",
actual: []string{"foo", "bar"},
cached: []string{"bar", "foo", "foobar"},
queued: []string{},
missing: []string{},
redundant: []string{"foobar"},
},
{
name: "redundant and queued values",
actual: []string{"foo", "bar"},
cached: []string{"foo", "foobar"},
queued: []string{"bar"},
missing: []string{},
redundant: []string{"foobar"},
},
{
name: "missing cached value",
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foo"},
queued: []string{},
missing: []string{"foobar"},
redundant: []string{},
},
{
name: "missing and queued values",
actual: []string{"foo", "bar", "foobar"},
cached: []string{"foo"},
queued: []string{"bar"},
missing: []string{"foobar"},
redundant: []string{},
},
{
name: "correct cache set",
actual: []string{"foo", "bar", "foobar"},
cached: []string{"bar", "foobar", "foo"},
queued: []string{},
missing: []string{},
redundant: []string{},
},
{
name: "queued cache value",
actual: []string{"foo", "bar", "foobar"},
cached: []string{"foobar", "foo"},
queued: []string{"bar"},
missing: []string{},
redundant: []string{},
},
}
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
testComparePods(test.actual, test.cached, test.queued, test.missing, test.redundant, t)
})
}
}
func testComparePods(actual, cached, queued, missing, redundant []string, t *testing.T) {
compare := CacheComparer{}
pods := []*v1.Pod{}
for _, uid := range actual {
pod := &v1.Pod{}
pod.UID = types.UID(uid)
pods = append(pods, pod)
}
queuedPods := []*v1.Pod{}
for _, uid := range queued {
pod := &v1.Pod{}
pod.UID = types.UID(uid)
queuedPods = append(queuedPods, pod)
}
nodeInfo := make(map[string]*schedulercache.NodeInfo)
for _, uid := range cached {
pod := &v1.Pod{}
pod.UID = types.UID(uid)
pod.Namespace = "ns"
pod.Name = uid
nodeInfo[uid] = schedulercache.NewNodeInfo(pod)
}
m, r := compare.ComparePods(pods, queuedPods, nodeInfo)
if !reflect.DeepEqual(m, missing) {
t.Errorf("missing expected to be %s; got %s", missing, m)
}
if !reflect.DeepEqual(r, redundant) {
t.Errorf("redundant expected to be %s; got %s", redundant, r)
}
}

View File

@ -0,0 +1,50 @@
/*
Copyright 2018 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 debugger
import (
corelisters "k8s.io/client-go/listers/core/v1"
internalcache "k8s.io/kubernetes/pkg/scheduler/internal/cache"
internalqueue "k8s.io/kubernetes/pkg/scheduler/internal/queue"
)
// CacheDebugger provides ways to check and write cache information for debugging.
type CacheDebugger struct {
Comparer CacheComparer
Dumper CacheDumper
}
// New creates a CacheDebugger.
func New(
nodeLister corelisters.NodeLister,
podLister corelisters.PodLister,
cache internalcache.Cache,
podQueue internalqueue.SchedulingQueue,
) *CacheDebugger {
return &CacheDebugger{
Comparer: CacheComparer{
NodeLister: nodeLister,
PodLister: podLister,
Cache: cache,
PodQueue: podQueue,
},
Dumper: CacheDumper{
cache: cache,
podQueue: podQueue,
},
}
}

Some files were not shown because too many files have changed in this diff Show More