Skip to content

K
k3s
Commit 2983c1d9 authored by David Oppenheimer's avatar David Oppenheimer
Browse files
Options

Merge pull request #22985 from Huawei-PaaS/inter-pod-affinity

Inter pod topological affinity and anti-affinity implementation
parents 28a8a234 82ba4f07
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 1588 additions and 152 deletions
cmd/integration/integration.go
View file @ 2983c1d9
......@@ -183,7 +183,7 @@ func startComponents(firstManifestURL, secondManifestURL string) (string, string
handler.delegate = m.Handler
// Scheduler
schedulerConfigFactory := factory.NewConfigFactory(cl, api.DefaultSchedulerName)
schedulerConfigFactory := factory.NewConfigFactory(cl, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
schedulerConfig, err := schedulerConfigFactory.Create()
if err != nil {
glog.Fatalf("Couldn't create scheduler config: %v", err)
......
docs/admin/kube-scheduler.md
View file @ 2983c1d9
......@@ -56,7 +56,9 @@ kube-scheduler
```
--address="0.0.0.0": The IP address to serve on (set to 0.0.0.0 for all interfaces)
--algorithm-provider="DefaultProvider": The scheduling algorithm provider to use, one of: DefaultProvider
--failure-domains="kubernetes.io/hostname,failure-domain.beta.kubernetes.io/zone,failure-domain.beta.kubernetes.io/region": Indicate the "all topologies" set for an empty topologyKey when it's used for PreferredDuringScheduling pod anti-affinity.
--google-json-key="": The Google Cloud Platform Service Account JSON Key to use for authentication.
--hard-pod-affinity-symmetric-weight=1: RequiredDuringScheduling affinity is not symmetric, but there is an implicit PreferredDuringScheduling affinity rule corresponding to every RequiredDuringScheduling affinity rule. --hard-pod-affinity-symmetric-weight represents the weight of implicit PreferredDuringScheduling affinity rule.
--kube-api-burst=100: Burst to use while talking with kubernetes apiserver
--kube-api-content-type="": ContentType of requests sent to apiserver. Passing application/vnd.kubernetes.protobuf is an experimental feature now.
--kube-api-qps=50: QPS to use while talking with kubernetes apiserver
......@@ -73,7 +75,7 @@ kube-scheduler
--scheduler-name="default-scheduler": Name of the scheduler, used to select which pods will be processed by this scheduler, based on pod's annotation with key 'scheduler.alpha.kubernetes.io/name'
```
###### Auto generated by spf13/cobra on 21-Apr-2016
###### Auto generated by spf13/cobra on 5-May-2016
<!-- BEGIN MUNGE: GENERATED_ANALYTICS -->
......
examples/examples_test.go
View file @ 2983c1d9
......@@ -333,7 +333,6 @@ func TestExampleObjectSchemas(t *testing.T) {
},
"../docs/user-guide/node-selection": {
"pod": &api.Pod{},
"pod-with-node-affinity": &api.Pod{},
},
"../examples/openshift-origin": {
"openshift-origin-namespace": &api.Namespace{},
......
hack/verify-flags/known-flags.txt
View file @ 2983c1d9
......@@ -129,6 +129,7 @@ experimental-prefix
external-hostname
external-ip
failover-timeout
failure-domains
fake-clientset
file-check-frequency
file-suffix
......@@ -153,6 +154,7 @@ google-json-key
grace-period
ha-domain
hairpin-mode
hard-pod-affinity-symmetric-weight
healthz-bind-address
healthz-port
horizontal-pod-autoscaler-sync-period
......
pkg/api/deep_copy_generated.go
View file @ 2983c1d9
......@@ -129,6 +129,9 @@ func init() {
DeepCopy_api_PersistentVolumeSpec,
DeepCopy_api_PersistentVolumeStatus,
DeepCopy_api_Pod,
DeepCopy_api_PodAffinity,
DeepCopy_api_PodAffinityTerm,
DeepCopy_api_PodAntiAffinity,
DeepCopy_api_PodAttachOptions,
DeepCopy_api_PodCondition,
DeepCopy_api_PodExecOptions,
......@@ -175,6 +178,7 @@ func init() {
DeepCopy_api_Volume,
DeepCopy_api_VolumeMount,
DeepCopy_api_VolumeSource,
DeepCopy_api_WeightedPodAffinityTerm,
); err != nil {
// if one of the deep copy functions is malformed, detect it immediately.
panic(err)
......@@ -199,6 +203,24 @@ func DeepCopy_api_Affinity(in Affinity, out *Affinity, c *conversion.Cloner) err
} else {
out.NodeAffinity = nil
}
if in.PodAffinity != nil {
in, out := in.PodAffinity, &out.PodAffinity
*out = new(PodAffinity)
if err := DeepCopy_api_PodAffinity(*in, *out, c); err != nil {
return err
}
} else {
out.PodAffinity = nil
}
if in.PodAntiAffinity != nil {
in, out := in.PodAntiAffinity, &out.PodAntiAffinity
*out = new(PodAntiAffinity)
if err := DeepCopy_api_PodAntiAffinity(*in, *out, c); err != nil {
return err
}
} else {
out.PodAntiAffinity = nil
}
return nil
}
......@@ -1964,6 +1986,79 @@ func DeepCopy_api_Pod(in Pod, out *Pod, c *conversion.Cloner) error {
return nil
}
func DeepCopy_api_PodAffinity(in PodAffinity, out *PodAffinity, c *conversion.Cloner) error {
if in.RequiredDuringSchedulingIgnoredDuringExecution != nil {
in, out := in.RequiredDuringSchedulingIgnoredDuringExecution, &out.RequiredDuringSchedulingIgnoredDuringExecution
*out = make([]PodAffinityTerm, len(in))
for i := range in {
if err := DeepCopy_api_PodAffinityTerm(in[i], &(*out)[i], c); err != nil {
return err
}
}
} else {
out.RequiredDuringSchedulingIgnoredDuringExecution = nil
}
if in.PreferredDuringSchedulingIgnoredDuringExecution != nil {
in, out := in.PreferredDuringSchedulingIgnoredDuringExecution, &out.PreferredDuringSchedulingIgnoredDuringExecution
*out = make([]WeightedPodAffinityTerm, len(in))
for i := range in {
if err := DeepCopy_api_WeightedPodAffinityTerm(in[i], &(*out)[i], c); err != nil {
return err
}
}
} else {
out.PreferredDuringSchedulingIgnoredDuringExecution = nil
}
return nil
}
func DeepCopy_api_PodAffinityTerm(in PodAffinityTerm, out *PodAffinityTerm, c *conversion.Cloner) error {
if in.LabelSelector != nil {
in, out := in.LabelSelector, &out.LabelSelector
*out = new(unversioned.LabelSelector)
if err := unversioned.DeepCopy_unversioned_LabelSelector(*in, *out, c); err != nil {
return err
}
} else {
out.LabelSelector = nil
}
if in.Namespaces != nil {
in, out := in.Namespaces, &out.Namespaces
*out = make([]string, len(in))
copy(*out, in)
} else {
out.Namespaces = nil
}
out.TopologyKey = in.TopologyKey
return nil
}
func DeepCopy_api_PodAntiAffinity(in PodAntiAffinity, out *PodAntiAffinity, c *conversion.Cloner) error {
if in.RequiredDuringSchedulingIgnoredDuringExecution != nil {
in, out := in.RequiredDuringSchedulingIgnoredDuringExecution, &out.RequiredDuringSchedulingIgnoredDuringExecution
*out = make([]PodAffinityTerm, len(in))
for i := range in {
if err := DeepCopy_api_PodAffinityTerm(in[i], &(*out)[i], c); err != nil {
return err
}
}
} else {
out.RequiredDuringSchedulingIgnoredDuringExecution = nil
}
if in.PreferredDuringSchedulingIgnoredDuringExecution != nil {
in, out := in.PreferredDuringSchedulingIgnoredDuringExecution, &out.PreferredDuringSchedulingIgnoredDuringExecution
*out = make([]WeightedPodAffinityTerm, len(in))
for i := range in {
if err := DeepCopy_api_WeightedPodAffinityTerm(in[i], &(*out)[i], c); err != nil {
return err
}
}
} else {
out.PreferredDuringSchedulingIgnoredDuringExecution = nil
}
return nil
}
func DeepCopy_api_PodAttachOptions(in PodAttachOptions, out *PodAttachOptions, c *conversion.Cloner) error {
if err := unversioned.DeepCopy_unversioned_TypeMeta(in.TypeMeta, &out.TypeMeta, c); err != nil {
return err
......@@ -3037,3 +3132,11 @@ func DeepCopy_api_VolumeSource(in VolumeSource, out *VolumeSource, c *conversion
}
return nil
}
func DeepCopy_api_WeightedPodAffinityTerm(in WeightedPodAffinityTerm, out *WeightedPodAffinityTerm, c *conversion.Cloner) error {
out.Weight = in.Weight
if err := DeepCopy_api_PodAffinityTerm(in.PodAffinityTerm, &out.PodAffinityTerm, c); err != nil {
return err
}
return nil
}
pkg/api/types.go
View file @ 2983c1d9
......@@ -1159,11 +1159,109 @@ const (
NodeSelectorOpLt NodeSelectorOperator = "Lt"
)
// Affinity is a group of affinity scheduling rules, currently
// only node affinity, but in the future also inter-pod affinity.
// Affinity is a group of affinity scheduling rules.
type Affinity struct {
// Describes node affinity scheduling rules for the pod.
NodeAffinity *NodeAffinity `json:"nodeAffinity,omitempty"`
// Describes pod affinity scheduling rules (e.g. co-locate this pod in the same node, zone, etc. as some other pod(s)).
PodAffinity *PodAffinity `json:"podAffinity,omitempty"`
// Describes pod anti-affinity scheduling rules (e.g. avoid putting this pod in the same node, zone, etc. as some other pod(s)).
PodAntiAffinity *PodAntiAffinity `json:"podAntiAffinity,omitempty"`
}
// Pod affinity is a group of inter pod affinity scheduling rules.
type PodAffinity struct {
// NOT YET IMPLEMENTED. TODO: Uncomment field once it is implemented.
// If the affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system will try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
// RequiredDuringSchedulingRequiredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingRequiredDuringExecution,omitempty"`
// If the affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system may or may not try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
RequiredDuringSchedulingIgnoredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingIgnoredDuringExecution,omitempty"`
// The scheduler will prefer to schedule pods to nodes that satisfy
// the affinity expressions specified by this field, but it may choose
// a node that violates one or more of the expressions. The node that is
// most preferred is the one with the greatest sum of weights, i.e.
// for each node that meets all of the scheduling requirements (resource
// request, requiredDuringScheduling affinity expressions, etc.),
// compute a sum by iterating through the elements of this field and adding
// "weight" to the sum if the node has pods which matches the corresponding podAffinityTerm; the
// node(s) with the highest sum are the most preferred.
PreferredDuringSchedulingIgnoredDuringExecution []WeightedPodAffinityTerm `json:"preferredDuringSchedulingIgnoredDuringExecution,omitempty"`
}
// Pod anti affinity is a group of inter pod anti affinity scheduling rules.
type PodAntiAffinity struct {
// NOT YET IMPLEMENTED. TODO: Uncomment field once it is implemented.
// If the anti-affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the anti-affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system will try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
// RequiredDuringSchedulingRequiredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingRequiredDuringExecution,omitempty"`
// If the anti-affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the anti-affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system may or may not try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
RequiredDuringSchedulingIgnoredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingIgnoredDuringExecution,omitempty"`
// The scheduler will prefer to schedule pods to nodes that satisfy
// the anti-affinity expressions specified by this field, but it may choose
// a node that violates one or more of the expressions. The node that is
// most preferred is the one with the greatest sum of weights, i.e.
// for each node that meets all of the scheduling requirements (resource
// request, requiredDuringScheduling anti-affinity expressions, etc.),
// compute a sum by iterating through the elements of this field and adding
// "weight" to the sum if the node has pods which matches the corresponding podAffinityTerm; the
// node(s) with the highest sum are the most preferred.
PreferredDuringSchedulingIgnoredDuringExecution []WeightedPodAffinityTerm `json:"preferredDuringSchedulingIgnoredDuringExecution,omitempty"`
}
// The weights of all of the matched WeightedPodAffinityTerm fields are added per-node to find the most preferred node(s)
type WeightedPodAffinityTerm struct {
// weight associated with matching the corresponding podAffinityTerm,
// in the range 1-100.
Weight int `json:"weight"`
// Required. A pod affinity term, associated with the corresponding weight.
PodAffinityTerm PodAffinityTerm `json:"podAffinityTerm"`
}
// Defines a set of pods (namely those matching the labelSelector
// relative to the given namespace(s)) that this pod should be
// co-located (affinity) or not co-located (anti-affinity) with,
// where co-located is defined as running on a node whose value of
// the label with key <topologyKey> matches that of any node on which
// a pod of the set of pods is running.
type PodAffinityTerm struct {
// A label query over a set of resources, in this case pods.
LabelSelector *unversioned.LabelSelector `json:"labelSelector,omitempty"`
// namespaces specifies which namespaces the labelSelector applies to (matches against);
// nil list means "this pod's namespace," empty list means "all namespaces"
// The json tag here is not "omitempty" since we need to distinguish nil and empty.
// See https://golang.org/pkg/encoding/json/#Marshal for more details.
Namespaces []string `json:"namespaces"`
// This pod should be co-located (affinity) or not co-located (anti-affinity) with the pods matching
// the labelSelector in the specified namespaces, where co-located is defined as running on a node
// whose value of the label with key topologyKey matches that of any node on which any of the
// selected pods is running.
// For PreferredDuringScheduling pod anti-affinity, empty topologyKey is interpreted as "all topologies"
// ("all topologies" here means all the topologyKeys indicated by scheduler command-line argument --failure-domains);
// for affinity and for RequiredDuringScheduling pod anti-affinity, empty topologyKey is not allowed.
TopologyKey string `json:"topologyKey,omitempty"`
}
// Node affinity is a group of node affinity scheduling rules.
......@@ -2604,4 +2702,14 @@ type RangeAllocation struct {
const (
// "default-scheduler" is the name of default scheduler.
DefaultSchedulerName = "default-scheduler"
// RequiredDuringScheduling affinity is not symmetric, but there is an implicit PreferredDuringScheduling affinity rule
// corresponding to every RequiredDuringScheduling affinity rule.
// When the --hard-pod-affinity-weight scheduler flag is not specified,
// DefaultHardPodAffinityWeight defines the weight of the implicit PreferredDuringScheduling affinity rule.
DefaultHardPodAffinitySymmetricWeight int = 1
// When the --failure-domains scheduler flag is not specified,
// DefaultFailureDomains defines the set of label keys used when TopologyKey is empty in PreferredDuringScheduling anti-affinity.
DefaultFailureDomains string = unversioned.LabelHostname + "," + unversioned.LabelZoneFailureDomain + "," + unversioned.LabelZoneRegion
)
pkg/api/unversioned/validation/validation.go
View file @ 2983c1d9
......@@ -17,17 +17,24 @@ limitations under the License.
package validation
import (
"fmt"
"k8s.io/kubernetes/pkg/api/unversioned"
apivalidation "k8s.io/kubernetes/pkg/api/validation"
"k8s.io/kubernetes/pkg/util/validation"
"k8s.io/kubernetes/pkg/util/validation/field"
)
var (
labelValueErrorMsg string = fmt.Sprintf(`must have at most %d characters, matching regex %s: e.g. "MyValue" or ""`, validation.LabelValueMaxLength, validation.LabelValueFmt)
qualifiedNameErrorMsg string = fmt.Sprintf(`must be a qualified name (at most %d characters, matching regex %s), with an optional DNS subdomain prefix (at most %d characters, matching regex %s) and slash (/): e.g. "MyName" or "example.com/MyName"`, validation.QualifiedNameMaxLength, validation.QualifiedNameFmt, validation.DNS1123SubdomainMaxLength, validation.DNS1123SubdomainFmt)
)
func ValidateLabelSelector(ps *unversioned.LabelSelector, fldPath *field.Path) field.ErrorList {
allErrs := field.ErrorList{}
if ps == nil {
return allErrs
}
allErrs = append(allErrs, apivalidation.ValidateLabels(ps.MatchLabels, fldPath.Child("matchLabels"))...)
allErrs = append(allErrs, ValidateLabels(ps.MatchLabels, fldPath.Child("matchLabels"))...)
for i, expr := range ps.MatchExpressions {
allErrs = append(allErrs, ValidateLabelSelectorRequirement(expr, fldPath.Child("matchExpressions").Index(i))...)
}
......@@ -48,6 +55,27 @@ func ValidateLabelSelectorRequirement(sr unversioned.LabelSelectorRequirement, f
default:
allErrs = append(allErrs, field.Invalid(fldPath.Child("operator"), sr.Operator, "not a valid selector operator"))
}
allErrs = append(allErrs, apivalidation.ValidateLabelName(sr.Key, fldPath.Child("key"))...)
allErrs = append(allErrs, ValidateLabelName(sr.Key, fldPath.Child("key"))...)
return allErrs
}
// ValidateLabelName validates that the label name is correctly defined.
func ValidateLabelName(labelName string, fldPath *field.Path) field.ErrorList {
allErrs := field.ErrorList{}
if !validation.IsQualifiedName(labelName) {
allErrs = append(allErrs, field.Invalid(fldPath, labelName, qualifiedNameErrorMsg))
}
return allErrs
}
// ValidateLabels validates that a set of labels are correctly defined.
func ValidateLabels(labels map[string]string, fldPath *field.Path) field.ErrorList {
allErrs := field.ErrorList{}
for k, v := range labels {
allErrs = append(allErrs, ValidateLabelName(k, fldPath)...)
if !validation.IsValidLabelValue(v) {
allErrs = append(allErrs, field.Invalid(fldPath, v, labelValueErrorMsg))
}
}
return allErrs
}
pkg/api/unversioned/validation/validation_test.go 0 → 100644
View file @ 2983c1d9
/*
Copyright 2016 The Kubernetes Authors All rights reserved.
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 (
"strings"
"testing"
"k8s.io/kubernetes/pkg/util/validation/field"
)
func TestValidateLabels(t *testing.T) {
successCases := []map[string]string{
{"simple": "bar"},
{"now-with-dashes": "bar"},
{"1-starts-with-num": "bar"},
{"1234": "bar"},
{"simple/simple": "bar"},
{"now-with-dashes/simple": "bar"},
{"now-with-dashes/now-with-dashes": "bar"},
{"now.with.dots/simple": "bar"},
{"now-with.dashes-and.dots/simple": "bar"},
{"1-num.2-num/3-num": "bar"},
{"1234/5678": "bar"},
{"1.2.3.4/5678": "bar"},
{"UpperCaseAreOK123": "bar"},
{"goodvalue": "123_-.BaR"},
}
for i := range successCases {
errs := ValidateLabels(successCases[i], field.NewPath("field"))
if len(errs) != 0 {
t.Errorf("case[%d] expected success, got %#v", i, errs)
}
}
labelNameErrorCases := []map[string]string{
{"nospecialchars^=@": "bar"},
{"cantendwithadash-": "bar"},
{"only/one/slash": "bar"},
{strings.Repeat("a", 254): "bar"},
}
for i := range labelNameErrorCases {
errs := ValidateLabels(labelNameErrorCases[i], field.NewPath("field"))
if len(errs) != 1 {
t.Errorf("case[%d] expected failure", i)
} else {
detail := errs[0].Detail
if detail != qualifiedNameErrorMsg {
t.Errorf("error detail %s should be equal %s", detail, qualifiedNameErrorMsg)
}
}
}
labelValueErrorCases := []map[string]string{
{"toolongvalue": strings.Repeat("a", 64)},
{"backslashesinvalue": "some\\bad\\value"},
{"nocommasallowed": "bad,value"},
{"strangecharsinvalue": "?#$notsogood"},
}
for i := range labelValueErrorCases {
errs := ValidateLabels(labelValueErrorCases[i], field.NewPath("field"))
if len(errs) != 1 {
t.Errorf("case[%d] expected failure", i)
} else {
detail := errs[0].Detail
if detail != labelValueErrorMsg {
t.Errorf("error detail %s should be equal %s", detail, labelValueErrorMsg)
}
}
}
}
pkg/api/unversioned/well_known_labels.go
View file @ 2983c1d9
......@@ -16,7 +16,12 @@ limitations under the License.
package unversioned
const LabelHostname = "kubernetes.io/hostname"
const LabelZoneFailureDomain = "failure-domain.beta.kubernetes.io/zone"
const LabelZoneRegion = "failure-domain.beta.kubernetes.io/region"
const LabelInstanceType = "beta.kubernetes.io/instance-type"
const (
// If you add a new topology domain here, also consider adding it to the set of default values
// for the scheduler's --failure-domain command-line argument.
LabelHostname = "kubernetes.io/hostname"
LabelZoneFailureDomain = "failure-domain.beta.kubernetes.io/zone"
LabelZoneRegion = "failure-domain.beta.kubernetes.io/region"
LabelInstanceType = "beta.kubernetes.io/instance-type"
)
pkg/api/v1/deep_copy_generated.go
View file @ 2983c1d9
......@@ -126,6 +126,9 @@ func init() {
DeepCopy_v1_PersistentVolumeSpec,
DeepCopy_v1_PersistentVolumeStatus,
DeepCopy_v1_Pod,
DeepCopy_v1_PodAffinity,
DeepCopy_v1_PodAffinityTerm,
DeepCopy_v1_PodAntiAffinity,
DeepCopy_v1_PodAttachOptions,
DeepCopy_v1_PodCondition,
DeepCopy_v1_PodExecOptions,
......@@ -172,6 +175,7 @@ func init() {
DeepCopy_v1_Volume,
DeepCopy_v1_VolumeMount,
DeepCopy_v1_VolumeSource,
DeepCopy_v1_WeightedPodAffinityTerm,
); err != nil {
// if one of the deep copy functions is malformed, detect it immediately.
panic(err)
......@@ -196,6 +200,24 @@ func DeepCopy_v1_Affinity(in Affinity, out *Affinity, c *conversion.Cloner) erro
} else {
out.NodeAffinity = nil
}
if in.PodAffinity != nil {
in, out := in.PodAffinity, &out.PodAffinity
*out = new(PodAffinity)
if err := DeepCopy_v1_PodAffinity(*in, *out, c); err != nil {
return err
}
} else {
out.PodAffinity = nil
}
if in.PodAntiAffinity != nil {
in, out := in.PodAntiAffinity, &out.PodAntiAffinity
*out = new(PodAntiAffinity)
if err := DeepCopy_v1_PodAntiAffinity(*in, *out, c); err != nil {
return err
}
} else {
out.PodAntiAffinity = nil
}
return nil
}
......@@ -1911,6 +1933,79 @@ func DeepCopy_v1_Pod(in Pod, out *Pod, c *conversion.Cloner) error {
return nil
}
func DeepCopy_v1_PodAffinity(in PodAffinity, out *PodAffinity, c *conversion.Cloner) error {
if in.RequiredDuringSchedulingIgnoredDuringExecution != nil {
in, out := in.RequiredDuringSchedulingIgnoredDuringExecution, &out.RequiredDuringSchedulingIgnoredDuringExecution
*out = make([]PodAffinityTerm, len(in))
for i := range in {
if err := DeepCopy_v1_PodAffinityTerm(in[i], &(*out)[i], c); err != nil {
return err
}
}
} else {
out.RequiredDuringSchedulingIgnoredDuringExecution = nil
}
if in.PreferredDuringSchedulingIgnoredDuringExecution != nil {
in, out := in.PreferredDuringSchedulingIgnoredDuringExecution, &out.PreferredDuringSchedulingIgnoredDuringExecution
*out = make([]WeightedPodAffinityTerm, len(in))
for i := range in {
if err := DeepCopy_v1_WeightedPodAffinityTerm(in[i], &(*out)[i], c); err != nil {
return err
}
}
} else {
out.PreferredDuringSchedulingIgnoredDuringExecution = nil
}
return nil
}
func DeepCopy_v1_PodAffinityTerm(in PodAffinityTerm, out *PodAffinityTerm, c *conversion.Cloner) error {
if in.LabelSelector != nil {
in, out := in.LabelSelector, &out.LabelSelector
*out = new(unversioned.LabelSelector)
if err := unversioned.DeepCopy_unversioned_LabelSelector(*in, *out, c); err != nil {
return err
}
} else {
out.LabelSelector = nil
}
if in.Namespaces != nil {
in, out := in.Namespaces, &out.Namespaces
*out = make([]string, len(in))
copy(*out, in)
} else {
out.Namespaces = nil
}
out.TopologyKey = in.TopologyKey
return nil
}
func DeepCopy_v1_PodAntiAffinity(in PodAntiAffinity, out *PodAntiAffinity, c *conversion.Cloner) error {
if in.RequiredDuringSchedulingIgnoredDuringExecution != nil {
in, out := in.RequiredDuringSchedulingIgnoredDuringExecution, &out.RequiredDuringSchedulingIgnoredDuringExecution
*out = make([]PodAffinityTerm, len(in))
for i := range in {
if err := DeepCopy_v1_PodAffinityTerm(in[i], &(*out)[i], c); err != nil {
return err
}
}
} else {
out.RequiredDuringSchedulingIgnoredDuringExecution = nil
}
if in.PreferredDuringSchedulingIgnoredDuringExecution != nil {
in, out := in.PreferredDuringSchedulingIgnoredDuringExecution, &out.PreferredDuringSchedulingIgnoredDuringExecution
*out = make([]WeightedPodAffinityTerm, len(in))
for i := range in {
if err := DeepCopy_v1_WeightedPodAffinityTerm(in[i], &(*out)[i], c); err != nil {
return err
}
}
} else {
out.PreferredDuringSchedulingIgnoredDuringExecution = nil
}
return nil
}
func DeepCopy_v1_PodAttachOptions(in PodAttachOptions, out *PodAttachOptions, c *conversion.Cloner) error {
if err := unversioned.DeepCopy_unversioned_TypeMeta(in.TypeMeta, &out.TypeMeta, c); err != nil {
return err
......@@ -2998,3 +3093,11 @@ func DeepCopy_v1_VolumeSource(in VolumeSource, out *VolumeSource, c *conversion.
}
return nil
}
func DeepCopy_v1_WeightedPodAffinityTerm(in WeightedPodAffinityTerm, out *WeightedPodAffinityTerm, c *conversion.Cloner) error {
out.Weight = in.Weight
if err := DeepCopy_v1_PodAffinityTerm(in.PodAffinityTerm, &out.PodAffinityTerm, c); err != nil {
return err
}
return nil
}
pkg/api/v1/generated.proto
View file @ 2983c1d9
......@@ -59,11 +59,16 @@ message AWSElasticBlockStoreVolumeSource {
optional bool readOnly = 4;
}
// Affinity is a group of affinity scheduling rules, currently
// only node affinity, but in the future also inter-pod affinity.
// Affinity is a group of affinity scheduling rules.
message Affinity {
// Describes node affinity scheduling rules for the pod.
optional NodeAffinity nodeAffinity = 1;
// Describes pod affinity scheduling rules (e.g. co-locate this pod in the same node, zone, etc. as some other pod(s)).
optional PodAffinity podAffinity = 2;
// Describes pod anti-affinity scheduling rules (e.g. avoid putting this pod in the same node, zone, etc. as some other pod(s)).
optional PodAntiAffinity podAntiAffinity = 3;
}
// AzureFile represents an Azure File Service mount on the host and bind mount to the pod.
......@@ -1726,6 +1731,96 @@ message Pod {
optional PodStatus status = 3;
}
// Pod affinity is a group of inter pod affinity scheduling rules.
message PodAffinity {
// NOT YET IMPLEMENTED. TODO: Uncomment field once it is implemented.
// If the affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system will try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
// RequiredDuringSchedulingRequiredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingRequiredDuringExecution,omitempty"`
// If the affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system may or may not try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
repeated PodAffinityTerm requiredDuringSchedulingIgnoredDuringExecution = 1;
// The scheduler will prefer to schedule pods to nodes that satisfy
// the affinity expressions specified by this field, but it may choose
// a node that violates one or more of the expressions. The node that is
// most preferred is the one with the greatest sum of weights, i.e.
// for each node that meets all of the scheduling requirements (resource
// request, requiredDuringScheduling affinity expressions, etc.),
// compute a sum by iterating through the elements of this field and adding
// "weight" to the sum if the node has pods which matches the corresponding podAffinityTerm; the
// node(s) with the highest sum are the most preferred.
repeated WeightedPodAffinityTerm preferredDuringSchedulingIgnoredDuringExecution = 2;
}
// Defines a set of pods (namely those matching the labelSelector
// relative to the given namespace(s)) that this pod should be
// co-located (affinity) or not co-located (anti-affinity) with,
// where co-located is defined as running on a node whose value of
// the label with key <topologyKey> tches that of any node on which
// a pod of the set of pods is running
message PodAffinityTerm {
// A label query over a set of resources, in this case pods.
optional k8s.io.kubernetes.pkg.api.unversioned.LabelSelector labelSelector = 1;
// namespaces specifies which namespaces the labelSelector applies to (matches against);
// nil list means "this pod's namespace," empty list means "all namespaces"
// The json tag here is not "omitempty" since we need to distinguish nil and empty.
// See https://golang.org/pkg/encoding/json/#Marshal for more details.
repeated string namespaces = 2;
// This pod should be co-located (affinity) or not co-located (anti-affinity) with the pods matching
// the labelSelector in the specified namespaces, where co-located is defined as running on a node
// whose value of the label with key topologyKey matches that of any node on which any of the
// selected pods is running.
// For PreferredDuringScheduling pod anti-affinity, empty topologyKey is interpreted as "all topologies"
// ("all topologies" here means all the topologyKeys indicated by scheduler command-line argument --failure-domains);
// for affinity and for RequiredDuringScheduling pod anti-affinity, empty topologyKey is not allowed.
optional string topologyKey = 3;
}
// Pod anti affinity is a group of inter pod anti affinity scheduling rules.
message PodAntiAffinity {
// NOT YET IMPLEMENTED. TODO: Uncomment field once it is implemented.
// If the anti-affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the anti-affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system will try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
// RequiredDuringSchedulingRequiredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingRequiredDuringExecution,omitempty"`
// If the anti-affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the anti-affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system may or may not try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
repeated PodAffinityTerm requiredDuringSchedulingIgnoredDuringExecution = 1;
// The scheduler will prefer to schedule pods to nodes that satisfy
// the anti-affinity expressions specified by this field, but it may choose
// a node that violates one or more of the expressions. The node that is
// most preferred is the one with the greatest sum of weights, i.e.
// for each node that meets all of the scheduling requirements (resource
// request, requiredDuringScheduling anti-affinity expressions, etc.),
// compute a sum by iterating through the elements of this field and adding
// "weight" to the sum if the node has pods which matches the corresponding podAffinityTerm; the
// node(s) with the highest sum are the most preferred.
repeated WeightedPodAffinityTerm preferredDuringSchedulingIgnoredDuringExecution = 2;
}
// PodAttachOptions is the query options to a Pod's remote attach call.
// ---
// TODO: merge w/ PodExecOptions below for stdin, stdout, etc
......@@ -2696,3 +2791,13 @@ message VolumeSource {
optional ConfigMapVolumeSource configMap = 19;
}
// The weights of all of the matched WeightedPodAffinityTerm fields are added per-node to find the most preferred node(s)
message WeightedPodAffinityTerm {
// weight associated with matching the corresponding podAffinityTerm,
// in the range 1-100.
optional int32 weight = 1;
// Required. A pod affinity term, associated with the corresponding weight.
optional PodAffinityTerm podAffinityTerm = 2;
}
pkg/api/v1/types.go
View file @ 2983c1d9
......@@ -1394,11 +1394,109 @@ const (
NodeSelectorOpLt NodeSelectorOperator = "Lt"
)
// Affinity is a group of affinity scheduling rules, currently
// only node affinity, but in the future also inter-pod affinity.
// Affinity is a group of affinity scheduling rules.
type Affinity struct {
// Describes node affinity scheduling rules for the pod.
NodeAffinity *NodeAffinity `json:"nodeAffinity,omitempty" protobuf:"bytes,1,opt,name=nodeAffinity"`
// Describes pod affinity scheduling rules (e.g. co-locate this pod in the same node, zone, etc. as some other pod(s)).
PodAffinity *PodAffinity `json:"podAffinity,omitempty" protobuf:"bytes,2,opt,name=podAffinity"`
// Describes pod anti-affinity scheduling rules (e.g. avoid putting this pod in the same node, zone, etc. as some other pod(s)).
PodAntiAffinity *PodAntiAffinity `json:"podAntiAffinity,omitempty" protobuf:"bytes,3,opt,name=podAntiAffinity"`
}
// Pod affinity is a group of inter pod affinity scheduling rules.
type PodAffinity struct {
// NOT YET IMPLEMENTED. TODO: Uncomment field once it is implemented.
// If the affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system will try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
// RequiredDuringSchedulingRequiredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingRequiredDuringExecution,omitempty"`
// If the affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system may or may not try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
RequiredDuringSchedulingIgnoredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingIgnoredDuringExecution,omitempty" protobuf:"bytes,1,rep,name=requiredDuringSchedulingIgnoredDuringExecution"`
// The scheduler will prefer to schedule pods to nodes that satisfy
// the affinity expressions specified by this field, but it may choose
// a node that violates one or more of the expressions. The node that is
// most preferred is the one with the greatest sum of weights, i.e.
// for each node that meets all of the scheduling requirements (resource
// request, requiredDuringScheduling affinity expressions, etc.),
// compute a sum by iterating through the elements of this field and adding
// "weight" to the sum if the node has pods which matches the corresponding podAffinityTerm; the
// node(s) with the highest sum are the most preferred.
PreferredDuringSchedulingIgnoredDuringExecution []WeightedPodAffinityTerm `json:"preferredDuringSchedulingIgnoredDuringExecution,omitempty" protobuf:"bytes,2,rep,name=preferredDuringSchedulingIgnoredDuringExecution"`
}
// Pod anti affinity is a group of inter pod anti affinity scheduling rules.
type PodAntiAffinity struct {
// NOT YET IMPLEMENTED. TODO: Uncomment field once it is implemented.
// If the anti-affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the anti-affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system will try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
// RequiredDuringSchedulingRequiredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingRequiredDuringExecution,omitempty"`
// If the anti-affinity requirements specified by this field are not met at
// scheduling time, the pod will not be scheduled onto the node.
// If the anti-affinity requirements specified by this field cease to be met
// at some point during pod execution (e.g. due to a pod label update), the
// system may or may not try to eventually evict the pod from its node.
// When there are multiple elements, the lists of nodes corresponding to each
// podAffinityTerm are intersected, i.e. all terms must be satisfied.
RequiredDuringSchedulingIgnoredDuringExecution []PodAffinityTerm `json:"requiredDuringSchedulingIgnoredDuringExecution,omitempty" protobuf:"bytes,1,rep,name=requiredDuringSchedulingIgnoredDuringExecution"`
// The scheduler will prefer to schedule pods to nodes that satisfy
// the anti-affinity expressions specified by this field, but it may choose
// a node that violates one or more of the expressions. The node that is
// most preferred is the one with the greatest sum of weights, i.e.
// for each node that meets all of the scheduling requirements (resource
// request, requiredDuringScheduling anti-affinity expressions, etc.),
// compute a sum by iterating through the elements of this field and adding
// "weight" to the sum if the node has pods which matches the corresponding podAffinityTerm; the
// node(s) with the highest sum are the most preferred.
PreferredDuringSchedulingIgnoredDuringExecution []WeightedPodAffinityTerm `json:"preferredDuringSchedulingIgnoredDuringExecution,omitempty" protobuf:"bytes,2,rep,name=preferredDuringSchedulingIgnoredDuringExecution"`
}
// The weights of all of the matched WeightedPodAffinityTerm fields are added per-node to find the most preferred node(s)
type WeightedPodAffinityTerm struct {
// weight associated with matching the corresponding podAffinityTerm,
// in the range 1-100.
Weight int32 `json:"weight" protobuf:"varint,1,opt,name=weight"`
// Required. A pod affinity term, associated with the corresponding weight.
PodAffinityTerm PodAffinityTerm `json:"podAffinityTerm" protobuf:"bytes,2,opt,name=podAffinityTerm"`
}
// Defines a set of pods (namely those matching the labelSelector
// relative to the given namespace(s)) that this pod should be
// co-located (affinity) or not co-located (anti-affinity) with,
// where co-located is defined as running on a node whose value of
// the label with key <topologyKey> tches that of any node on which
// a pod of the set of pods is running
type PodAffinityTerm struct {
// A label query over a set of resources, in this case pods.
LabelSelector *unversioned.LabelSelector `json:"labelSelector,omitempty" protobuf:"bytes,1,opt,name=labelSelector"`
// namespaces specifies which namespaces the labelSelector applies to (matches against);
// nil list means "this pod's namespace," empty list means "all namespaces"
// The json tag here is not "omitempty" since we need to distinguish nil and empty.
// See https://golang.org/pkg/encoding/json/#Marshal for more details.
Namespaces []string `json:"namespaces" protobuf:"bytes,2,rep,name=namespaces"`
// This pod should be co-located (affinity) or not co-located (anti-affinity) with the pods matching
// the labelSelector in the specified namespaces, where co-located is defined as running on a node
// whose value of the label with key topologyKey matches that of any node on which any of the
// selected pods is running.
// For PreferredDuringScheduling pod anti-affinity, empty topologyKey is interpreted as "all topologies"
// ("all topologies" here means all the topologyKeys indicated by scheduler command-line argument --failure-domains);
// for affinity and for RequiredDuringScheduling pod anti-affinity, empty topologyKey is not allowed.
TopologyKey string `json:"topologyKey,omitempty" protobuf:"bytes,3,opt,name=topologyKey"`
}
// Node affinity is a group of node affinity scheduling rules.
......
pkg/api/v1/types_swagger_doc_generated.go
View file @ 2983c1d9
......@@ -40,8 +40,10 @@ func (AWSElasticBlockStoreVolumeSource) SwaggerDoc() map[string]string {
}
var map_Affinity = map[string]string{
"": "Affinity is a group of affinity scheduling rules, currently only node affinity, but in the future also inter-pod affinity.",
"nodeAffinity": "Describes node affinity scheduling rules for the pod.",
"": "Affinity is a group of affinity scheduling rules.",
"nodeAffinity": "Describes node affinity scheduling rules for the pod.",
"podAffinity": "Describes pod affinity scheduling rules (e.g. co-locate this pod in the same node, zone, etc. as some other pod(s)).",
"podAntiAffinity": "Describes pod anti-affinity scheduling rules (e.g. avoid putting this pod in the same node, zone, etc. as some other pod(s)).",
}
func (Affinity) SwaggerDoc() map[string]string {
......@@ -1086,6 +1088,37 @@ func (Pod) SwaggerDoc() map[string]string {
return map_Pod
}
var map_PodAffinity = map[string]string{
"": "Pod affinity is a group of inter pod affinity scheduling rules.",
"requiredDuringSchedulingIgnoredDuringExecution": "NOT YET IMPLEMENTED. TODO: Uncomment field once it is implemented. If the affinity requirements specified by this field are not met at scheduling time, the pod will not be scheduled onto the node. If the affinity requirements specified by this field cease to be met at some point during pod execution (e.g. due to a pod label update), the system will try to eventually evict the pod from its node. When there are multiple elements, the lists of nodes corresponding to each podAffinityTerm are intersected, i.e. all terms must be satisfied. RequiredDuringSchedulingRequiredDuringExecution []PodAffinityTerm `json:\"requiredDuringSchedulingRequiredDuringExecution,omitempty\"` If the affinity requirements specified by this field are not met at scheduling time, the pod will not be scheduled onto the node. If the affinity requirements specified by this field cease to be met at some point during pod execution (e.g. due to a pod label update), the system may or may not try to eventually evict the pod from its node. When there are multiple elements, the lists of nodes corresponding to each podAffinityTerm are intersected, i.e. all terms must be satisfied.",
"preferredDuringSchedulingIgnoredDuringExecution": "The scheduler will prefer to schedule pods to nodes that satisfy the affinity expressions specified by this field, but it may choose a node that violates one or more of the expressions. The node that is most preferred is the one with the greatest sum of weights, i.e. for each node that meets all of the scheduling requirements (resource request, requiredDuringScheduling affinity expressions, etc.), compute a sum by iterating through the elements of this field and adding \"weight\" to the sum if the node has pods which matches the corresponding podAffinityTerm; the node(s) with the highest sum are the most preferred.",
}
func (PodAffinity) SwaggerDoc() map[string]string {
return map_PodAffinity
}
var map_PodAffinityTerm = map[string]string{
"": "Defines a set of pods (namely those matching the labelSelector relative to the given namespace(s)) that this pod should be co-located (affinity) or not co-located (anti-affinity) with, where co-located is defined as running on a node whose value of the label with key <topologyKey> tches that of any node on which a pod of the set of pods is running",
"labelSelector": "A label query over a set of resources, in this case pods.",
"namespaces": "namespaces specifies which namespaces the labelSelector applies to (matches against); nil list means \"this pod's namespace,\" empty list means \"all namespaces\" The json tag here is not \"omitempty\" since we need to distinguish nil and empty. See https://golang.org/pkg/encoding/json/#Marshal for more details.",
"topologyKey": "This pod should be co-located (affinity) or not co-located (anti-affinity) with the pods matching the labelSelector in the specified namespaces, where co-located is defined as running on a node whose value of the label with key topologyKey matches that of any node on which any of the selected pods is running. For PreferredDuringScheduling pod anti-affinity, empty topologyKey is interpreted as \"all topologies\" (\"all topologies\" here means all the topologyKeys indicated by scheduler command-line argument --failure-domains); for affinity and for RequiredDuringScheduling pod anti-affinity, empty topologyKey is not allowed.",
}
func (PodAffinityTerm) SwaggerDoc() map[string]string {
return map_PodAffinityTerm
}
var map_PodAntiAffinity = map[string]string{
"": "Pod anti affinity is a group of inter pod anti affinity scheduling rules.",
"requiredDuringSchedulingIgnoredDuringExecution": "NOT YET IMPLEMENTED. TODO: Uncomment field once it is implemented. If the anti-affinity requirements specified by this field are not met at scheduling time, the pod will not be scheduled onto the node. If the anti-affinity requirements specified by this field cease to be met at some point during pod execution (e.g. due to a pod label update), the system will try to eventually evict the pod from its node. When there are multiple elements, the lists of nodes corresponding to each podAffinityTerm are intersected, i.e. all terms must be satisfied. RequiredDuringSchedulingRequiredDuringExecution []PodAffinityTerm `json:\"requiredDuringSchedulingRequiredDuringExecution,omitempty\"` If the anti-affinity requirements specified by this field are not met at scheduling time, the pod will not be scheduled onto the node. If the anti-affinity requirements specified by this field cease to be met at some point during pod execution (e.g. due to a pod label update), the system may or may not try to eventually evict the pod from its node. When there are multiple elements, the lists of nodes corresponding to each podAffinityTerm are intersected, i.e. all terms must be satisfied.",
"preferredDuringSchedulingIgnoredDuringExecution": "The scheduler will prefer to schedule pods to nodes that satisfy the anti-affinity expressions specified by this field, but it may choose a node that violates one or more of the expressions. The node that is most preferred is the one with the greatest sum of weights, i.e. for each node that meets all of the scheduling requirements (resource request, requiredDuringScheduling anti-affinity expressions, etc.), compute a sum by iterating through the elements of this field and adding \"weight\" to the sum if the node has pods which matches the corresponding podAffinityTerm; the node(s) with the highest sum are the most preferred.",
}
func (PodAntiAffinity) SwaggerDoc() map[string]string {
return map_PodAntiAffinity
}
var map_PodAttachOptions = map[string]string{
"": "PodAttachOptions is the query options to a Pod's remote attach call.",
"stdin": "Stdin if true, redirects the standard input stream of the pod for this call. Defaults to false.",
......@@ -1628,4 +1661,14 @@ func (VolumeSource) SwaggerDoc() map[string]string {
return map_VolumeSource
}
var map_WeightedPodAffinityTerm = map[string]string{
"": "The weights of all of the matched WeightedPodAffinityTerm fields are added per-node to find the most preferred node(s)",
"weight": "weight associated with matching the corresponding podAffinityTerm, in the range 1-100.",
"podAffinityTerm": "Required. A pod affinity term, associated with the corresponding weight.",
}
func (WeightedPodAffinityTerm) SwaggerDoc() map[string]string {
return map_WeightedPodAffinityTerm
}
// AUTO-GENERATED FUNCTIONS END HERE
pkg/apis/apps/validation/validation.go
View file @ 2983c1d9
......@@ -56,7 +56,7 @@ func ValidatePodTemplateSpecForPetSet(template *api.PodTemplateSpec, selector la
// fail. We should really check that the union of the given volumes and volumeClaims match
// volume mounts in the containers.
// allErrs = append(allErrs, apivalidation.ValidatePodTemplateSpec(template, fldPath)...)
allErrs = append(allErrs, apivalidation.ValidateLabels(template.Labels, fldPath.Child("labels"))...)
allErrs = append(allErrs, unversionedvalidation.ValidateLabels(template.Labels, fldPath.Child("labels"))...)
allErrs = append(allErrs, apivalidation.ValidateAnnotations(template.Annotations, fldPath.Child("annotations"))...)
allErrs = append(allErrs, apivalidation.ValidatePodSpecificAnnotations(template.Annotations, fldPath.Child("annotations"))...)
}
......
pkg/apis/componentconfig/deep_copy_generated.go
View file @ 2983c1d9
......@@ -189,6 +189,8 @@ func DeepCopy_componentconfig_KubeSchedulerConfiguration(in KubeSchedulerConfigu
out.KubeAPIQPS = in.KubeAPIQPS
out.KubeAPIBurst = in.KubeAPIBurst
out.SchedulerName = in.SchedulerName
out.HardPodAffinitySymmetricWeight = in.HardPodAffinitySymmetricWeight
out.FailureDomains = in.FailureDomains
if err := DeepCopy_componentconfig_LeaderElectionConfiguration(in.LeaderElection, &out.LeaderElection, c); err != nil {
return err
}
......
pkg/apis/componentconfig/types.go
View file @ 2983c1d9
......@@ -372,6 +372,12 @@ type KubeSchedulerConfiguration struct {
// will be processed by this scheduler, based on pod's annotation with
// key 'scheduler.alpha.kubernetes.io/name'.
SchedulerName string `json:"schedulerName"`
// 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 int `json:"hardPodAffinitySymmetricWeight"`
// Indicate the "all topologies" set for empty topologyKey when it's used for PreferredDuringScheduling pod anti-affinity.
FailureDomains string `json:"failureDomains"`
// leaderElection defines the configuration of leader election client.
LeaderElection LeaderElectionConfiguration `json:"leaderElection"`
}
......
pkg/apis/componentconfig/v1alpha1/conversion_generated.go
View file @ 2983c1d9
......@@ -163,6 +163,8 @@ func autoConvert_v1alpha1_KubeSchedulerConfiguration_To_componentconfig_KubeSche
out.KubeAPIQPS = in.KubeAPIQPS
out.KubeAPIBurst = int32(in.KubeAPIBurst)
out.SchedulerName = in.SchedulerName
out.HardPodAffinitySymmetricWeight = in.HardPodAffinitySymmetricWeight
out.FailureDomains = in.FailureDomains
if err := Convert_v1alpha1_LeaderElectionConfiguration_To_componentconfig_LeaderElectionConfiguration(&in.LeaderElection, &out.LeaderElection, s); err != nil {
return err
}
......@@ -191,6 +193,8 @@ func autoConvert_componentconfig_KubeSchedulerConfiguration_To_v1alpha1_KubeSche
out.KubeAPIQPS = in.KubeAPIQPS
out.KubeAPIBurst = int(in.KubeAPIBurst)
out.SchedulerName = in.SchedulerName
out.HardPodAffinitySymmetricWeight = in.HardPodAffinitySymmetricWeight
out.FailureDomains = in.FailureDomains
if err := Convert_componentconfig_LeaderElectionConfiguration_To_v1alpha1_LeaderElectionConfiguration(&in.LeaderElection, &out.LeaderElection, s); err != nil {
return err
}
......
pkg/apis/componentconfig/v1alpha1/deep_copy_generated.go
View file @ 2983c1d9
......@@ -98,6 +98,8 @@ func DeepCopy_v1alpha1_KubeSchedulerConfiguration(in KubeSchedulerConfiguration,
out.KubeAPIQPS = in.KubeAPIQPS
out.KubeAPIBurst = in.KubeAPIBurst
out.SchedulerName = in.SchedulerName
out.HardPodAffinitySymmetricWeight = in.HardPodAffinitySymmetricWeight
out.FailureDomains = in.FailureDomains
if err := DeepCopy_v1alpha1_LeaderElectionConfiguration(in.LeaderElection, &out.LeaderElection, c); err != nil {
return err
}
......
pkg/apis/componentconfig/v1alpha1/defaults.go
View file @ 2983c1d9
......@@ -89,6 +89,12 @@ func SetDefaults_KubeSchedulerConfiguration(obj *KubeSchedulerConfiguration) {
if obj.SchedulerName == "" {
obj.SchedulerName = api.DefaultSchedulerName
}
if obj.HardPodAffinitySymmetricWeight == 0 {
obj.HardPodAffinitySymmetricWeight = api.DefaultHardPodAffinitySymmetricWeight
}
if obj.FailureDomains == "" {
obj.FailureDomains = api.DefaultFailureDomains
}
}
func SetDefaults_LeaderElectionConfiguration(obj *LeaderElectionConfiguration) {
......
pkg/apis/componentconfig/v1alpha1/types.go
View file @ 2983c1d9
......@@ -105,6 +105,12 @@ type KubeSchedulerConfiguration struct {
// will be processed by this scheduler, based on pod's annotation with
// key 'scheduler.alpha.kubernetes.io/name'.
SchedulerName string `json:"schedulerName"`
// 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 int `json:"hardPodAffinitySymmetricWeight"`
// Indicate the "all topologies" set for empty topologyKey when it's used for PreferredDuringScheduling pod anti-affinity.
FailureDomains string `json:"failureDomains"`
// leaderElection defines the configuration of leader election client.
LeaderElection LeaderElectionConfiguration `json:"leaderElection"`
}
......
plugin/cmd/kube-scheduler/app/options/options.go
View file @ 2983c1d9
......@@ -67,5 +67,9 @@ func (s *SchedulerServer) AddFlags(fs *pflag.FlagSet) {
fs.Float32Var(&s.KubeAPIQPS, "kube-api-qps", s.KubeAPIQPS, "QPS to use while talking with kubernetes apiserver")
fs.Int32Var(&s.KubeAPIBurst, "kube-api-burst", s.KubeAPIBurst, "Burst to use while talking with kubernetes apiserver")
fs.StringVar(&s.SchedulerName, "scheduler-name", s.SchedulerName, "Name of the scheduler, used to select which pods will be processed by this scheduler, based on pod's annotation with key 'scheduler.alpha.kubernetes.io/name'")
fs.IntVar(&s.HardPodAffinitySymmetricWeight, "hard-pod-affinity-symmetric-weight", api.DefaultHardPodAffinitySymmetricWeight,
"RequiredDuringScheduling affinity is not symmetric, but there is an implicit PreferredDuringScheduling affinity rule corresponding "+
"to every RequiredDuringScheduling affinity rule. --hard-pod-affinity-symmetric-weight represents the weight of implicit PreferredDuringScheduling affinity rule.")
fs.StringVar(&s.FailureDomains, "failure-domains", api.DefaultFailureDomains, "Indicate the \"all topologies\" set for an empty topologyKey when it's used for PreferredDuringScheduling pod anti-affinity.")
leaderelection.BindFlags(&s.LeaderElection, fs)
}
plugin/cmd/kube-scheduler/app/server.go
View file @ 2983c1d9
......@@ -107,7 +107,7 @@ func Run(s *options.SchedulerServer) error {
glog.Fatal(server.ListenAndServe())
}()
configFactory := factory.NewConfigFactory(kubeClient, s.SchedulerName)
configFactory := factory.NewConfigFactory(kubeClient, s.SchedulerName, s.HardPodAffinitySymmetricWeight, s.FailureDomains)
config, err := createConfig(s, configFactory)
if err != nil {
......
plugin/pkg/scheduler/algorithm/predicates/error.go
View file @ 2983c1d9
......@@ -30,6 +30,7 @@ var (
ErrDiskConflict = newPredicateFailureError("NoDiskConflict")
ErrVolumeZoneConflict = newPredicateFailureError("NoVolumeZoneConflict")
ErrNodeSelectorNotMatch = newPredicateFailureError("MatchNodeSelector")
ErrPodAffinityNotMatch = newPredicateFailureError("MatchInterPodAffinity")
ErrPodNotMatchHostName = newPredicateFailureError("HostName")
ErrPodNotFitsHostPorts = newPredicateFailureError("PodFitsHostPorts")
ErrNodeLabelPresenceViolated = newPredicateFailureError("CheckNodeLabelPresence")
......
plugin/pkg/scheduler/algorithm/priorities/interpod_affinity.go 0 → 100644
View file @ 2983c1d9
/*
Copyright 2016 The Kubernetes Authors All rights reserved.
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 (
"github.com/golang/glog"
"k8s.io/kubernetes/pkg/api"
"k8s.io/kubernetes/pkg/labels"
"k8s.io/kubernetes/plugin/pkg/scheduler/algorithm"
"k8s.io/kubernetes/plugin/pkg/scheduler/algorithm/predicates"
priorityutil "k8s.io/kubernetes/plugin/pkg/scheduler/algorithm/priorities/util"
schedulerapi "k8s.io/kubernetes/plugin/pkg/scheduler/api"
"k8s.io/kubernetes/plugin/pkg/scheduler/schedulercache"
)
type InterPodAffinity struct {
info predicates.NodeInfo
nodeLister algorithm.NodeLister
podLister algorithm.PodLister
hardPodAffinityWeight int
failureDomains priorityutil.Topologies
}
func NewInterPodAffinityPriority(info predicates.NodeInfo, nodeLister algorithm.NodeLister, podLister algorithm.PodLister, hardPodAffinityWeight int, failureDomains []string) algorithm.PriorityFunction {
interPodAffinity := &InterPodAffinity{
info: info,
nodeLister: nodeLister,
podLister: podLister,
hardPodAffinityWeight: hardPodAffinityWeight,
failureDomains: priorityutil.Topologies{DefaultKeys: failureDomains},
}
return interPodAffinity.CalculateInterPodAffinityPriority
}
// countPodsThatMatchPodAffinityTerm counts the number of given pods that match the podAffinityTerm.
func (ipa *InterPodAffinity) CountPodsThatMatchPodAffinityTerm(pod *api.Pod, podsForMatching []*api.Pod, node *api.Node, podAffinityTerm api.PodAffinityTerm) (int, error) {
matchedCount := 0
for _, ep := range podsForMatching {
match, err := ipa.failureDomains.CheckIfPodMatchPodAffinityTerm(ep, pod, podAffinityTerm,
func(ep *api.Pod) (*api.Node, error) {
return ipa.info.GetNodeInfo(ep.Spec.NodeName)
},
func(pod *api.Pod) (*api.Node, error) {
return node, nil
},
)
if err != nil {
return 0, err
}
if match {
matchedCount++
}
}
return matchedCount, nil
}
// CountWeightByPodMatchAffinityTerm counts the weight to topologyCounts for all the given pods that match the podAffinityTerm.
func (ipa *InterPodAffinity) CountWeightByPodMatchAffinityTerm(pod *api.Pod, podsForMatching []*api.Pod, weight int, podAffinityTerm api.PodAffinityTerm, node *api.Node) (int, error) {
if weight == 0 {
return 0, nil
}
// get the pods which are there in that particular node
podsMatchedCount, err := ipa.CountPodsThatMatchPodAffinityTerm(pod, podsForMatching, node, podAffinityTerm)
return weight * podsMatchedCount, err
}
// compute a sum by iterating through the elements of weightedPodAffinityTerm and adding
// "weight" to the sum if the corresponding PodAffinityTerm is satisfied for
// that node; the node(s) with the highest sum are the most preferred.
// Symmetry need to be considered for preferredDuringSchedulingIgnoredDuringExecution from podAffinity & podAntiAffinity,
// symmetry need to be considered for hard requirements from podAffinity
func (ipa *InterPodAffinity) CalculateInterPodAffinityPriority(pod *api.Pod, nodeNameToInfo map[string]*schedulercache.NodeInfo, nodeLister algorithm.NodeLister) (schedulerapi.HostPriorityList, error) {
nodes, err := nodeLister.List()
if err != nil {
return nil, err
}
allPods, err := ipa.podLister.List(labels.Everything())
if err != nil {
return nil, err
}
affinity, err := api.GetAffinityFromPodAnnotations(pod.Annotations)
if err != nil {
return nil, err
}
// convert the topology key based weights to the node name based weights
var maxCount int
var minCount int
counts := map[string]int{}
for _, node := range nodes.Items {
totalCount := 0
// count weights for the weighted pod affinity
if affinity.PodAffinity != nil {
for _, weightedTerm := range affinity.PodAffinity.PreferredDuringSchedulingIgnoredDuringExecution {
weightedCount, err := ipa.CountWeightByPodMatchAffinityTerm(pod, allPods, weightedTerm.Weight, weightedTerm.PodAffinityTerm, &node)
if err != nil {
return nil, err
}
totalCount += weightedCount
}
}
// count weights for the weighted pod anti-affinity
if affinity.PodAntiAffinity != nil {
for _, weightedTerm := range affinity.PodAntiAffinity.PreferredDuringSchedulingIgnoredDuringExecution {
weightedCount, err := ipa.CountWeightByPodMatchAffinityTerm(pod, allPods, (0 - weightedTerm.Weight), weightedTerm.PodAffinityTerm, &node)
if err != nil {
return nil, err
}
totalCount += weightedCount
}
}
// reverse direction checking: count weights for the inter-pod affinity/anti-affinity rules
// that are indicated by existing pods on the node.
for _, ep := range allPods {
epAffinity, err := api.GetAffinityFromPodAnnotations(ep.Annotations)
if err != nil {
return nil, err
}
if epAffinity.PodAffinity != nil {
// count the implicit weight for the hard pod affinity indicated by the existing pod.
if ipa.hardPodAffinityWeight > 0 {
var podAffinityTerms []api.PodAffinityTerm
if len(epAffinity.PodAffinity.RequiredDuringSchedulingIgnoredDuringExecution) != 0 {
podAffinityTerms = epAffinity.PodAffinity.RequiredDuringSchedulingIgnoredDuringExecution
}
// TODO: Uncomment this block when implement RequiredDuringSchedulingRequiredDuringExecution.
//if len(affinity.PodAffinity.RequiredDuringSchedulingRequiredDuringExecution) != 0 {
// podAffinityTerms = append(podAffinityTerms, affinity.PodAffinity.RequiredDuringSchedulingRequiredDuringExecution...)
//}
for _, epAffinityTerm := range podAffinityTerms {
match, err := ipa.failureDomains.CheckIfPodMatchPodAffinityTerm(pod, ep, epAffinityTerm,
func(pod *api.Pod) (*api.Node, error) { return &node, nil },
func(ep *api.Pod) (*api.Node, error) { return ipa.info.GetNodeInfo(ep.Spec.NodeName) },
)
if err != nil {
return nil, err
}
if match {
totalCount += ipa.hardPodAffinityWeight
}
}
}
// count weight for the weighted pod affinity indicated by the existing pod.
for _, epWeightedTerm := range epAffinity.PodAffinity.PreferredDuringSchedulingIgnoredDuringExecution {
match, err := ipa.failureDomains.CheckIfPodMatchPodAffinityTerm(pod, ep, epWeightedTerm.PodAffinityTerm,
func(pod *api.Pod) (*api.Node, error) { return &node, nil },
func(ep *api.Pod) (*api.Node, error) { return ipa.info.GetNodeInfo(ep.Spec.NodeName) },
)
if err != nil {
return nil, err
}
if match {
totalCount += epWeightedTerm.Weight
}
}
}
// count weight for the weighted pod anti-affinity indicated by the existing pod.
if epAffinity.PodAntiAffinity != nil {
for _, epWeightedTerm := range epAffinity.PodAntiAffinity.PreferredDuringSchedulingIgnoredDuringExecution {
match, err := ipa.failureDomains.CheckIfPodMatchPodAffinityTerm(pod, ep, epWeightedTerm.PodAffinityTerm,
func(pod *api.Pod) (*api.Node, error) { return &node, nil },
func(ep *api.Pod) (*api.Node, error) { return ipa.info.GetNodeInfo(ep.Spec.NodeName) },
)
if err != nil {
return nil, err
}
if match {
totalCount -= epWeightedTerm.Weight
}
}
}
}
counts[node.Name] = totalCount
if counts[node.Name] > maxCount {
maxCount = counts[node.Name]
}
if counts[node.Name] < minCount {
minCount = counts[node.Name]
}
}
// calculate final priority score for each node
result := []schedulerapi.HostPriority{}
for _, node := range nodes.Items {
fScore := float64(0)
if (maxCount - minCount) > 0 {
fScore = 10 * (float64(counts[node.Name]-minCount) / float64(maxCount-minCount))
}
result = append(result, schedulerapi.HostPriority{Host: node.Name, Score: int(fScore)})
glog.V(10).Infof(
"%v -> %v: InterPodAffinityPriority, Score: (%d)", pod.Name, node.Name, int(fScore),
)
}
return result, nil
}
plugin/pkg/scheduler/algorithm/priorities/util/non_zero.go
View file @ 2983c1d9
......@@ -16,7 +16,12 @@ limitations under the License.
package util
import "k8s.io/kubernetes/pkg/api"
import (
"k8s.io/kubernetes/pkg/api"
"k8s.io/kubernetes/pkg/api/unversioned"
"k8s.io/kubernetes/pkg/labels"
"k8s.io/kubernetes/pkg/util/sets"
)
// For each of these resources, a pod that doesn't request the resource explicitly
// will be treated as having requested the amount indicated below, for the purpose
......@@ -48,3 +53,83 @@ func GetNonzeroRequests(requests *api.ResourceList) (int64, int64) {
}
return outMilliCPU, outMemory
}
// FilterPodsByNameSpaces filters the pods based the given list of namespaces,
// empty set of namespaces means all namespaces.
func FilterPodsByNameSpaces(names sets.String, pods []*api.Pod) []*api.Pod {
if len(pods) == 0 || len(names) == 0 {
return pods
}
result := []*api.Pod{}
for _, pod := range pods {
if names.Has(pod.Namespace) {
result = append(result, pod)
}
}
return result
}
// GetNamespacesFromPodAffinityTerm returns a set of names
// according to the namespaces indicated in podAffinityTerm.
// if the NameSpaces is nil considers the given pod's namespace
// if the Namespaces is empty list then considers all the namespaces
func GetNamespacesFromPodAffinityTerm(pod *api.Pod, podAffinityTerm api.PodAffinityTerm) sets.String {
names := sets.String{}
if podAffinityTerm.Namespaces == nil {
names.Insert(pod.Namespace)
} else if len(podAffinityTerm.Namespaces) != 0 {
names.Insert(podAffinityTerm.Namespaces...)
}
return names
}
// NodesHaveSameTopologyKeyInternal checks if nodeA and nodeB have same label value with given topologyKey as label key.
func NodesHaveSameTopologyKeyInternal(nodeA, nodeB *api.Node, topologyKey string) bool {
return nodeA.Labels != nil && nodeB.Labels != nil && len(nodeA.Labels[topologyKey]) > 0 && nodeA.Labels[topologyKey] == nodeB.Labels[topologyKey]
}
type Topologies struct {
DefaultKeys []string
}
// NodesHaveSameTopologyKey checks if nodeA and nodeB have same label value with given topologyKey as label key.
// If the topologyKey is nil/empty, check if the two nodes have any of the default topologyKeys, and have same corresponding label value.
func (tps *Topologies) NodesHaveSameTopologyKey(nodeA *api.Node, nodeB *api.Node, topologyKey string) bool {
if len(topologyKey) == 0 {
// assumes this is allowed only for PreferredDuringScheduling pod anti-affinity (ensured by api/validation)
for _, defaultKey := range tps.DefaultKeys {
if NodesHaveSameTopologyKeyInternal(nodeA, nodeB, defaultKey) {
return true
}
}
return false
} else {
return NodesHaveSameTopologyKeyInternal(nodeA, nodeB, topologyKey)
}
}
type getNodeFunc func(*api.Pod) (*api.Node, error)
// CheckIfPodMatchPodAffinityTerm checks if podB's affinity request is compatible with podA
func (tps *Topologies) CheckIfPodMatchPodAffinityTerm(podA *api.Pod, podB *api.Pod, podBAffinityTerm api.PodAffinityTerm, getNodeA, getNodeB getNodeFunc) (bool, error) {
names := GetNamespacesFromPodAffinityTerm(podB, podBAffinityTerm)
if len(names) != 0 && !names.Has(podA.Namespace) {
return false, nil
}
labelSelector, err := unversioned.LabelSelectorAsSelector(podBAffinityTerm.LabelSelector)
if err != nil || !labelSelector.Matches(labels.Set(podA.Labels)) {
return false, err
}
podANode, err := getNodeA(podA)
if err != nil {
return false, err
}
podBNode, err := getNodeB(podB)
if err != nil {
return false, err
}
return tps.NodesHaveSameTopologyKey(podANode, podBNode, podBAffinityTerm.TopologyKey), nil
}
plugin/pkg/scheduler/algorithmprovider/defaults/compatibility_test.go
View file @ 2983c1d9
......@@ -22,6 +22,7 @@ import (
"net/http/httptest"
"k8s.io/kubernetes/pkg/api"
"k8s.io/kubernetes/pkg/api/testapi"
"k8s.io/kubernetes/pkg/client/restclient"
client "k8s.io/kubernetes/pkg/client/unversioned"
......@@ -116,7 +117,7 @@ func TestCompatibility_v1_Scheduler(t *testing.T) {
defer server.Close()
client := client.NewOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
if _, err := factory.NewConfigFactory(client, "some-scheduler-name").CreateFromConfig(policy); err != nil {
if _, err := factory.NewConfigFactory(client, "some-scheduler-name", api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains).CreateFromConfig(policy); err != nil {
t.Errorf("%s: Error constructing: %v", v, err)
continue
}
......
plugin/pkg/scheduler/algorithmprovider/defaults/defaults.go
View file @ 2983c1d9
......@@ -32,8 +32,10 @@ import (
"github.com/golang/glog"
)
// GCE instances can have up to 16 PD volumes attached.
const DefaultMaxGCEPDVolumes = 16
const (
// GCE instances can have up to 16 PD volumes attached.
DefaultMaxGCEPDVolumes = 16
)
// getMaxVols checks the max PD volumes environment variable, otherwise returning a default value
func getMaxVols(defaultVal int) int {
......@@ -71,7 +73,7 @@ func init() {
},
)
// PodFitsPorts has been replaced by PodFitsHostPorts for better user understanding.
// For backwards compatibility with 1.0, PodFitsPorts is regitered as well.
// For backwards compatibility with 1.0, PodFitsPorts is registered as well.
factory.RegisterFitPredicate("PodFitsPorts", predicates.PodFitsHostPorts)
// 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
......@@ -125,6 +127,13 @@ func defaultPredicates() sets.String {
// GeneralPredicates are the predicates that are enforced by all Kubernetes components
// (e.g. kubelet and all schedulers)
factory.RegisterFitPredicate("GeneralPredicates", predicates.GeneralPredicates),
// Fit is determined by inter-pod affinity.
factory.RegisterFitPredicateFactory(
"MatchInterPodAffinity",
func(args factory.PluginFactoryArgs) algorithm.FitPredicate {
return predicates.NewPodAffinityPredicate(args.NodeInfo, args.PodLister, args.FailureDomains)
},
),
)
}
......@@ -153,5 +162,16 @@ func defaultPriorities() sets.String {
Weight: 1,
},
),
//pods should be placed in the same topological domain (e.g. same node, same rack, same zone, same power domain, etc.)
//as some other pods, or, conversely, should not be placed in the same topological domain as some other pods.
factory.RegisterPriorityConfigFactory(
"InterPodAffinityPriority",
factory.PriorityConfigFactory{
Function: func(args factory.PluginFactoryArgs) algorithm.PriorityFunction {
return priorities.NewInterPodAffinityPriority(args.NodeInfo, args.NodeLister, args.PodLister, args.HardPodAffinitySymmetricWeight, args.FailureDomains)
},
Weight: 1,
},
),
)
}
plugin/pkg/scheduler/factory/factory.go
View file @ 2983c1d9
......@@ -21,6 +21,7 @@ package factory
import (
"fmt"
"math/rand"
"strings"
"sync"
"sync/atomic"
"time"
......@@ -34,6 +35,7 @@ import (
"k8s.io/kubernetes/pkg/types"
"k8s.io/kubernetes/pkg/util/runtime"
"k8s.io/kubernetes/pkg/util/sets"
utilvalidation "k8s.io/kubernetes/pkg/util/validation"
"k8s.io/kubernetes/plugin/pkg/scheduler"
"k8s.io/kubernetes/plugin/pkg/scheduler/algorithm"
"k8s.io/kubernetes/plugin/pkg/scheduler/algorithm/predicates"
......@@ -83,10 +85,18 @@ type ConfigFactory struct {
// processed by this scheduler, based on pods's annotation key:
// 'scheduler.alpha.kubernetes.io/name'
SchedulerName string
// 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 int
// Indicate the "all topologies" set for empty topologyKey when it's used for PreferredDuringScheduling pod anti-affinity.
FailureDomains string
}
// Initializes the factory.
func NewConfigFactory(client *client.Client, schedulerName string) *ConfigFactory {
func NewConfigFactory(client *client.Client, schedulerName string, hardPodAffinitySymmetricWeight int, failureDomains string) *ConfigFactory {
stopEverything := make(chan struct{})
schedulerCache := schedulercache.New(30*time.Second, stopEverything)
......@@ -95,15 +105,17 @@ func NewConfigFactory(client *client.Client, schedulerName string) *ConfigFactor
PodQueue: cache.NewFIFO(cache.MetaNamespaceKeyFunc),
ScheduledPodLister: &cache.StoreToPodLister{},
// Only nodes in the "Ready" condition with status == "True" are schedulable
NodeLister: &cache.StoreToNodeLister{},
PVLister: &cache.StoreToPVFetcher{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
PVCLister: &cache.StoreToPVCFetcher{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
ServiceLister: &cache.StoreToServiceLister{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
ControllerLister: &cache.StoreToReplicationControllerLister{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
ReplicaSetLister: &cache.StoreToReplicaSetLister{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
schedulerCache: schedulerCache,
StopEverything: stopEverything,
SchedulerName: schedulerName,
NodeLister: &cache.StoreToNodeLister{},
PVLister: &cache.StoreToPVFetcher{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
PVCLister: &cache.StoreToPVCFetcher{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
ServiceLister: &cache.StoreToServiceLister{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
ControllerLister: &cache.StoreToReplicationControllerLister{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
ReplicaSetLister: &cache.StoreToReplicaSetLister{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
schedulerCache: schedulerCache,
StopEverything: stopEverything,
SchedulerName: schedulerName,
HardPodAffinitySymmetricWeight: hardPodAffinitySymmetricWeight,
FailureDomains: failureDomains,
}
c.PodLister = schedulerCache
......@@ -287,6 +299,18 @@ func (f *ConfigFactory) CreateFromConfig(policy schedulerapi.Policy) (*scheduler
// Creates a scheduler from a set of registered fit predicate keys and priority keys.
func (f *ConfigFactory) CreateFromKeys(predicateKeys, priorityKeys sets.String, extenders []algorithm.SchedulerExtender) (*scheduler.Config, error) {
glog.V(2).Infof("creating scheduler with fit predicates '%v' and priority functions '%v", predicateKeys, priorityKeys)
if f.HardPodAffinitySymmetricWeight < 0 || f.HardPodAffinitySymmetricWeight > 100 {
return nil, fmt.Errorf("invalid hardPodAffinitySymmetricWeight: %d, must be in the range 0-100", f.HardPodAffinitySymmetricWeight)
}
failureDomainArgs := strings.Split(f.FailureDomains, ",")
for _, failureDomain := range failureDomainArgs {
if !utilvalidation.IsQualifiedName(failureDomain) {
return nil, fmt.Errorf("invalid failure domain: %s", failureDomain)
}
}
pluginArgs := PluginFactoryArgs{
PodLister: f.PodLister,
ServiceLister: f.ServiceLister,
......@@ -297,6 +321,8 @@ func (f *ConfigFactory) CreateFromKeys(predicateKeys, priorityKeys sets.String,
NodeInfo: &predicates.CachedNodeInfo{StoreToNodeLister: f.NodeLister},
PVInfo: f.PVLister,
PVCInfo: f.PVCLister,
HardPodAffinitySymmetricWeight: f.HardPodAffinitySymmetricWeight,
FailureDomains: sets.NewString(failureDomainArgs...).List(),
}
predicateFuncs, err := getFitPredicateFunctions(predicateKeys, pluginArgs)
if err != nil {
......
plugin/pkg/scheduler/factory/factory_test.go
View file @ 2983c1d9
......@@ -47,7 +47,7 @@ func TestCreate(t *testing.T) {
server := httptest.NewServer(&handler)
defer server.Close()
client := client.NewOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
factory := NewConfigFactory(client, api.DefaultSchedulerName)
factory := NewConfigFactory(client, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
factory.Create()
}
......@@ -65,7 +65,7 @@ func TestCreateFromConfig(t *testing.T) {
server := httptest.NewServer(&handler)
defer server.Close()
client := client.NewOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
factory := NewConfigFactory(client, api.DefaultSchedulerName)
factory := NewConfigFactory(client, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
// Pre-register some predicate and priority functions
RegisterFitPredicate("PredicateOne", PredicateOne)
......@@ -106,7 +106,7 @@ func TestCreateFromEmptyConfig(t *testing.T) {
server := httptest.NewServer(&handler)
defer server.Close()
client := client.NewOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
factory := NewConfigFactory(client, api.DefaultSchedulerName)
factory := NewConfigFactory(client, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
configData = []byte(`{}`)
if err := runtime.DecodeInto(latestschedulerapi.Codec, configData, &policy); err != nil {
......@@ -148,7 +148,7 @@ func TestDefaultErrorFunc(t *testing.T) {
mux.Handle(testapi.Default.ResourcePath("pods", "bar", "foo"), &handler)
server := httptest.NewServer(mux)
defer server.Close()
factory := NewConfigFactory(client.NewOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}}), api.DefaultSchedulerName)
factory := NewConfigFactory(client.NewOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}}), api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
queue := cache.NewFIFO(cache.MetaNamespaceKeyFunc)
podBackoff := podBackoff{
perPodBackoff: map[types.NamespacedName]*backoffEntry{},
......@@ -318,9 +318,9 @@ func TestResponsibleForPod(t *testing.T) {
defer server.Close()
client := client.NewOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
// factory of "default-scheduler"
factoryDefaultScheduler := NewConfigFactory(client, api.DefaultSchedulerName)
factoryDefaultScheduler := NewConfigFactory(client, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
// factory of "foo-scheduler"
factoryFooScheduler := NewConfigFactory(client, "foo-scheduler")
factoryFooScheduler := NewConfigFactory(client, "foo-scheduler", api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
// scheduler annotaions to be tested
schedulerAnnotationFitsDefault := map[string]string{"scheduler.alpha.kubernetes.io/name": "default-scheduler"}
schedulerAnnotationFitsFoo := map[string]string{"scheduler.alpha.kubernetes.io/name": "foo-scheduler"}
......@@ -370,3 +370,63 @@ func TestResponsibleForPod(t *testing.T) {
}
}
}
func TestInvalidHardPodAffinitySymmetricWeight(t *testing.T) {
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
// TODO: Uncomment when fix #19254
// defer server.Close()
client := client.NewOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
// factory of "default-scheduler"
factory := NewConfigFactory(client, api.DefaultSchedulerName, -1, api.DefaultFailureDomains)
_, err := factory.Create()
if err == nil {
t.Errorf("expected err: invalid hardPodAffinitySymmetricWeight, got nothing")
}
}
func TestInvalidFactoryArgs(t *testing.T) {
handler := utiltesting.FakeHandler{
StatusCode: 500,
ResponseBody: "",
T: t,
}
server := httptest.NewServer(&handler)
defer server.Close()
client := client.NewOrDie(&restclient.Config{Host: server.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
testCases := []struct {
hardPodAffinitySymmetricWeight int
failureDomains string
expectErr string
}{
{
hardPodAffinitySymmetricWeight: -1,
failureDomains: api.DefaultFailureDomains,
expectErr: "invalid hardPodAffinitySymmetricWeight: -1, must be in the range 0-100",
},
{
hardPodAffinitySymmetricWeight: 101,
failureDomains: api.DefaultFailureDomains,
expectErr: "invalid hardPodAffinitySymmetricWeight: 101, must be in the range 0-100",
},
{
hardPodAffinitySymmetricWeight: 0,
failureDomains: "INVALID_FAILURE_DOMAINS",
expectErr: "invalid failure domain: INVALID_FAILURE_DOMAINS",
},
}
for _, test := range testCases {
factory := NewConfigFactory(client, api.DefaultSchedulerName, test.hardPodAffinitySymmetricWeight, test.failureDomains)
_, err := factory.Create()
if err == nil {
t.Errorf("expected err: %s, got nothing", test.expectErr)
}
}
}
plugin/pkg/scheduler/factory/plugins.go
View file @ 2983c1d9
......@@ -33,14 +33,16 @@ import (
// PluginFactoryArgs are passed to all plugin factory functions.
type PluginFactoryArgs struct {
PodLister algorithm.PodLister
ServiceLister algorithm.ServiceLister
ControllerLister algorithm.ControllerLister
ReplicaSetLister algorithm.ReplicaSetLister
NodeLister algorithm.NodeLister
NodeInfo predicates.NodeInfo
PVInfo predicates.PersistentVolumeInfo
PVCInfo predicates.PersistentVolumeClaimInfo
PodLister algorithm.PodLister
ServiceLister algorithm.ServiceLister
ControllerLister algorithm.ControllerLister
ReplicaSetLister algorithm.ReplicaSetLister
NodeLister algorithm.NodeLister
NodeInfo predicates.NodeInfo
PVInfo predicates.PersistentVolumeInfo
PVCInfo predicates.PersistentVolumeClaimInfo
HardPodAffinitySymmetricWeight int
FailureDomains []string
}
// A FitPredicateFactory produces a FitPredicate from the given args.
......
test/component/scheduler/perf/util.go
View file @ 2983c1d9
......@@ -62,7 +62,7 @@ func mustSetupScheduler() (schedulerConfigFactory *factory.ConfigFactory, destro
Burst: 5000,
})
schedulerConfigFactory = factory.NewConfigFactory(c, api.DefaultSchedulerName)
schedulerConfigFactory = factory.NewConfigFactory(c, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
schedulerConfig, err := schedulerConfigFactory.Create()
if err != nil {
panic("Couldn't create scheduler config")
......
test/e2e/node-selection/pod-with-pod-affinity.yaml 0 → 100644
View file @ 2983c1d9
apiVersion: v1
kind: Pod
metadata:
name: with-newlabels
annotations:
scheduler.alpha.kubernetes.io/affinity: >
{
"podAffinity": {
"requiredDuringSchedulingIgnoredDuringExecution": [{
"labelSelector": {
"matchExpressions": [{
"key": "security",
"operator": "In",
"values":["S1"]
}]
},
"topologyKey": "kubernetes.io/hostname"
}]
},
"podAntiAffinity": {
"requiredDuringSchedulingIgnoredDuringExecution": [{
"labelSelector": {
"matchExpressions": [{
"key": "security",
"operator": "In",
"values":["S2"]
}]
},
"topologyKey": "kubernetes.io/hostname"
}]
}
}
another-annotation-key: another-annotation-value
spec:
containers:
- name: with-newlabels
image: gcr.io/google_containers/pause:2.0
test/integration/extender_test.go
View file @ 2983c1d9
......@@ -241,7 +241,7 @@ func TestSchedulerExtender(t *testing.T) {
}
policy.APIVersion = testapi.Default.GroupVersion().String()
schedulerConfigFactory := factory.NewConfigFactory(restClient, api.DefaultSchedulerName)
schedulerConfigFactory := factory.NewConfigFactory(restClient, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
schedulerConfig, err := schedulerConfigFactory.CreateFromConfig(policy)
if err != nil {
t.Fatalf("Couldn't create scheduler config: %v", err)
......
test/integration/scheduler_test.go
View file @ 2983c1d9
......@@ -68,7 +68,7 @@ func TestUnschedulableNodes(t *testing.T) {
restClient := client.NewOrDie(&restclient.Config{Host: s.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
schedulerConfigFactory := factory.NewConfigFactory(restClient, api.DefaultSchedulerName)
schedulerConfigFactory := factory.NewConfigFactory(restClient, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
schedulerConfig, err := schedulerConfigFactory.Create()
if err != nil {
t.Fatalf("Couldn't create scheduler config: %v", err)
......@@ -315,7 +315,7 @@ func TestMultiScheduler(t *testing.T) {
// 1. create and start default-scheduler
restClient := client.NewOrDie(&restclient.Config{Host: s.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
schedulerConfigFactory := factory.NewConfigFactory(restClient, api.DefaultSchedulerName)
schedulerConfigFactory := factory.NewConfigFactory(restClient, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
schedulerConfig, err := schedulerConfigFactory.Create()
if err != nil {
t.Fatalf("Couldn't create scheduler config: %v", err)
......@@ -386,7 +386,7 @@ func TestMultiScheduler(t *testing.T) {
// 5. create and start a scheduler with name "foo-scheduler"
restClient2 := client.NewOrDie(&restclient.Config{Host: s.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
schedulerConfigFactory2 := factory.NewConfigFactory(restClient2, "foo-scheduler")
schedulerConfigFactory2 := factory.NewConfigFactory(restClient2, "foo-scheduler", api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
schedulerConfig2, err := schedulerConfigFactory2.Create()
if err != nil {
t.Errorf("Couldn't create scheduler config: %v", err)
......@@ -484,7 +484,7 @@ func TestAllocatable(t *testing.T) {
// 1. create and start default-scheduler
restClient := client.NewOrDie(&restclient.Config{Host: s.URL, ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
schedulerConfigFactory := factory.NewConfigFactory(restClient, api.DefaultSchedulerName)
schedulerConfigFactory := factory.NewConfigFactory(restClient, api.DefaultSchedulerName, api.DefaultHardPodAffinitySymmetricWeight, api.DefaultFailureDomains)
schedulerConfig, err := schedulerConfigFactory.Create()
if err != nil {
t.Fatalf("Couldn't create scheduler config: %v", err)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment