[stable/coredns] Add optional cluster-proportional-autoscaler (#20359)

* [stable/coredns] add optional cluster-proportional-autoscaler

Signed-off-by: Alex Williams <alex.williams@skyscanner.net>

* [stable/coredns] autoscaler config map created before

Signed-off-by: Alex Williams <alex.williams@skyscanner.net>

* [stable/coredns] split into separate files. Update README.

Signed-off-by: Alex Williams <alex.williams@skyscanner.net>

* [stable/coredns] ensure RBAC is enabled when creating autoscaler role

Signed-off-by: Alex Williams <alex.williams@skyscanner.net>
This commit is contained in:
Alex Williams
2020-01-29 02:54:14 -08:00
committed by GitHub
parent 9637a42ebd
commit b69d37a5bd
9 changed files with 270 additions and 5 deletions
+1 -1
View File
@@ -1,6 +1,6 @@
apiVersion: v1
name: coredns
version: 1.8.5
version: 1.9.0
appVersion: 1.6.6
description: CoreDNS is a DNS server that chains plugins and provides Kubernetes DNS Services
keywords:
+32 -2
View File
@@ -67,14 +67,29 @@ The command removes all the Kubernetes components associated with the chart and
| `rbac.create` | If true, create & use RBAC resources | true |
| `rbac.pspEnable` | Specifies whether a PodSecurityPolicy should be created. | `false` |
| `isClusterService` | Specifies whether chart should be deployed as cluster-service or normal k8s app. | true |
| `priorityClassName` | Name of Priority Class to assign pods | `nil` |
| `priorityClassName` | Name of Priority Class to assign pods | `""` |
| `servers` | Configuration for CoreDNS and plugins | See values.yml |
| `affinity` | Affinity settings for pod assignment | {} |
| `nodeSelector` | Node labels for pod assignment | {} |
| `tolerations` | Tolerations for pod assignment | [] |
| `zoneFiles` | Configure custom Zone files | [] |
| `extraSecrets` | Optional array of secrets to mount inside the CoreDNS container | [] |
| `customLabels` | Optional labels for Deployment, Pod, Service, ServiceMonitor objects | {} |
| `customLabels` | Optional labels for Deployment(s), Pod, Service, ServiceMonitor objects | {} |
| `autoscaler.enabled` | Optionally enabled a cluster-proportional-autoscaler for CoreDNS | `false` |
| `autoscaler.coresPerReplica` | Number of cores in the cluster per CoreDNS replica | `256` |
| `autoscaler.nodesPerReplica` | Number of nodes in the cluster per CoreDNS replica | `16` |
| `autoscaler.image.repository` | The image repository to pull autoscaler from | k8s.gcr.io/cluster-proportional-autoscaler-amd64 |
| `autoscaler.image.tag` | The image tag to pull autoscaler from | `1.7.1` |
| `autoscaler.image.pullPolicy` | Image pull policy for the autoscaler | IfNotPresent |
| `autoscaler.priorityClassName` | Optional priority class for the autoscaler pod. `priorityClassName` used if not set. | `""` |
| `autoscaler.affinity` | Affinity settings for pod assignment for autoscaler | {} |
| `autoscaler.nodeSelector` | Node labels for pod assignment for autoscaler | {} |
| `autoscaler.tolerations` | Tolerations for pod assignment for autoscaler | [] |
| `autoscaler.resources.limits.cpu` | Container maximum CPU for cluster-proportional-autoscaler | `20m` |
| `autoscaler.resources.limits.memory` | Container maximum memory for cluster-proportional-autoscaler | `10Mi` |
| `autoscaler.resources.requests.cpu` | Container requested CPU for cluster-proportional-autoscaler | `20m` |
| `autoscaler.resources.requests.memory` | Container requested memory for cluster-proportional-autoscaler | `10Mi` |
See `values.yaml` for configuration notes. Specify each parameter using the `--set key=value[,key=value]` argument to `helm install`. For example,
```console
@@ -104,3 +119,18 @@ create external loadbalancers with both "TCP" and "UDP" protocols. So
When deploying CoreDNS with `serviceType="LoadBalancer"` on such cloud
environments, make sure you do not attempt to use both protocols at the same
time.
## Autoscaling
By setting `autoscaler.enabled = true` a
[cluster-proportional-autoscaler](https://github.com/kubernetes-incubator/cluster-proportional-autoscaler)
will be deployed. This will default to a coredns replica for every 256 cores, or
16 nodes in the cluster. These can be changed with `autoscaler.coresPerReplica`
and `autoscaler.nodesPerReplica`. When cluster is using large nodes (with more
cores), `coresPerReplica` should dominate. If using small nodes,
`nodesPerReplica` should dominate.
This also creates a ServiceAccount, ClusterRole, and ClusterRoleBinding for
the autoscaler deployment.
`replicaCount` is ignored if this is enabled.
@@ -0,0 +1,35 @@
{{- if and .Values.autoscaler.enabled .Values.rbac.create }}
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
name: {{ template "coredns.fullname" . }}-autoscaler
labels:
app.kubernetes.io/managed-by: {{ .Release.Service | quote }}
app.kubernetes.io/instance: {{ .Release.Name | quote }}
helm.sh/chart: "{{ .Chart.Name }}-{{ .Chart.Version | replace "+" "_" }}"
{{- if .Values.isClusterService }}
k8s-app: {{ .Chart.Name }}-autoscaler
kubernetes.io/cluster-service: "true"
kubernetes.io/name: "CoreDNS"
{{- end }}
app.kubernetes.io/name: {{ template "coredns.name" . }}-autoscaler
{{- if .Values.customLabels }}
{{ toYaml .Values.customLabels | indent 4 }}
{{- end }}
rules:
- apiGroups: [""]
resources: ["nodes"]
verbs: ["list","watch"]
- apiGroups: [""]
resources: ["replicationcontrollers/scale"]
verbs: ["get", "update"]
- apiGroups: ["extensions", "apps"]
resources: ["deployments/scale", "replicasets/scale"]
verbs: ["get", "update"]
# Remove the configmaps rule once below issue is fixed:
# kubernetes-incubator/cluster-proportional-autoscaler#16
- apiGroups: [""]
resources: ["configmaps"]
verbs: ["get", "create"]
{{- end }}
@@ -0,0 +1,28 @@
{{- if and .Values.autoscaler.enabled .Values.rbac.create }}
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: {{ template "coredns.fullname" . }}-autoscaler
labels:
app.kubernetes.io/managed-by: {{ .Release.Service | quote }}
app.kubernetes.io/instance: {{ .Release.Name | quote }}
helm.sh/chart: "{{ .Chart.Name }}-{{ .Chart.Version | replace "+" "_" }}"
{{- if .Values.isClusterService }}
k8s-app: {{ .Chart.Name }}-autoscaler
kubernetes.io/cluster-service: "true"
kubernetes.io/name: "CoreDNS"
{{- end }}
app.kubernetes.io/name: {{ template "coredns.name" . }}-autoscaler
{{- if .Values.customLabels }}
{{ toYaml .Values.customLabels | indent 4 }}
{{- end }}
roleRef:
apiGroup: rbac.authorization.k8s.io
kind: ClusterRole
name: {{ template "coredns.fullname" . }}-autoscaler
subjects:
- kind: ServiceAccount
name: {{ template "coredns.fullname" . }}-autoscaler
namespace: {{ .Release.Namespace }}
{{- end }}
@@ -0,0 +1,30 @@
{{- if .Values.autoscaler.enabled }}
---
kind: ConfigMap
apiVersion: v1
metadata:
name: {{ template "coredns.fullname" . }}-autoscaler
namespace: {{ .Release.Namespace }}
labels:
app.kubernetes.io/managed-by: {{ .Release.Service | quote }}
app.kubernetes.io/instance: {{ .Release.Name | quote }}
helm.sh/chart: "{{ .Chart.Name }}-{{ .Chart.Version | replace "+" "_" }}"
{{- if .Values.isClusterService }}
k8s-app: {{ .Chart.Name }}-autoscaler
kubernetes.io/cluster-service: "true"
kubernetes.io/name: "CoreDNS"
{{- end }}
app.kubernetes.io/name: {{ template "coredns.name" . }}-autoscaler
{{- if .Values.customLabels }}
{{ toYaml .Values.customLabels | indent 4 }}
{{- end }}
data:
# When cluster is using large nodes(with more cores), "coresPerReplica" should dominate.
# If using small nodes, "nodesPerReplica" should dominate.
linear: |-
{
"coresPerReplica": {{ .Values.autoscaler.coresPerReplica | float64 }},
"nodesPerReplica": {{ .Values.autoscaler.nodesPerReplica | float64 }},
"preventSinglePointFailure": true
}
{{- end }}
@@ -0,0 +1,76 @@
{{- if .Values.autoscaler.enabled }}
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: {{ template "coredns.fullname" . }}-autoscaler
namespace: {{ .Release.Namespace }}
labels:
app.kubernetes.io/managed-by: {{ .Release.Service | quote }}
app.kubernetes.io/instance: {{ .Release.Name | quote }}
helm.sh/chart: "{{ .Chart.Name }}-{{ .Chart.Version | replace "+" "_" }}"
{{- if .Values.isClusterService }}
k8s-app: {{ .Chart.Name }}-autoscaler
kubernetes.io/cluster-service: "true"
kubernetes.io/name: "CoreDNS"
{{- end }}
app.kubernetes.io/name: {{ template "coredns.name" . }}-autoscaler
{{- if .Values.customLabels }}
{{ toYaml .Values.customLabels | indent 4 }}
{{- end }}
spec:
selector:
matchLabels:
app.kubernetes.io/instance: {{ .Release.Name | quote }}
{{- if .Values.isClusterService }}
k8s-app: {{ .Chart.Name }}-autoscaler
{{- end }}
app.kubernetes.io/name: {{ template "coredns.name" . }}-autoscaler
template:
metadata:
labels:
{{- if .Values.isClusterService }}
k8s-app: {{ .Chart.Name }}-autoscaler
{{- end }}
app.kubernetes.io/name: {{ template "coredns.name" . }}-autoscaler
app.kubernetes.io/instance: {{ .Release.Name | quote }}
{{- if .Values.customLabels }}
{{ toYaml .Values.customLabels | nindent 8 }}
{{- end }}
annotations:
{{- if .Values.isClusterService }}
scheduler.alpha.kubernetes.io/critical-pod: ''
scheduler.alpha.kubernetes.io/tolerations: '[{"key":"CriticalAddonsOnly", "operator":"Exists"}]'
{{- end }}
spec:
serviceAccountName: {{ template "coredns.fullname" . }}-autoscaler
{{- $priorityClassName := default .Values.priorityClassName .Values.autoscaler.priorityClassName }}
{{- if $priorityClassName }}
priorityClassName: {{ $priorityClassName | quote }}
{{- end }}
{{- if .Values.autoscaler.affinity }}
affinity:
{{ toYaml .Values.autoscaler.affinity | indent 8 }}
{{- end }}
{{- if .Values.autoscaler.tolerations }}
tolerations:
{{ toYaml .Values.autoscaler.tolerations | indent 8 }}
{{- end }}
{{- if .Values.autoscaler.nodeSelector }}
nodeSelector:
{{ toYaml .Values.autoscaler.nodeSelector | indent 8 }}
{{- end }}
containers:
- name: autoscaler
image: "{{ .Values.autoscaler.image.repository }}:{{ .Values.autoscaler.image.tag }}"
imagePullPolicy: {{ .Values.autoscaler.image.pullPolicy }}
resources:
{{ toYaml .Values.autoscaler.resources | indent 10 }}
command:
- /cluster-proportional-autoscaler
- --namespace={{ .Release.Namespace }}
- --configmap={{ template "coredns.fullname" . }}-autoscaler
- --target=Deployment/{{ template "coredns.fullname" . }}
- --logtostderr=true
- --v=2
{{- end }}
+7
View File
@@ -16,7 +16,14 @@ metadata:
{{ toYaml .Values.customLabels | indent 4 }}
{{- end }}
spec:
{{- if not .Values.autoscaler.enabled }}
replicas: {{ .Values.replicaCount }}
{{- end }}
strategy:
type: RollingUpdate
rollingUpdate:
maxUnavailable: 1
maxSurge: 10%
selector:
matchLabels:
app.kubernetes.io/instance: {{ .Release.Name | quote }}
@@ -0,0 +1,21 @@
{{- if .Values.autoscaler.enabled }}
---
apiVersion: v1
kind: ServiceAccount
metadata:
name: {{ template "coredns.fullname" . }}-autoscaler
namespace: {{ .Release.Namespace }}
labels:
app.kubernetes.io/managed-by: {{ .Release.Service | quote }}
app.kubernetes.io/instance: {{ .Release.Name | quote }}
helm.sh/chart: "{{ .Chart.Name }}-{{ .Chart.Version | replace "+" "_" }}"
{{- if .Values.isClusterService }}
k8s-app: {{ .Chart.Name }}-autoscaler
kubernetes.io/cluster-service: "true"
kubernetes.io/name: "CoreDNS"
{{- end }}
app.kubernetes.io/name: {{ template "coredns.name" . }}-autoscaler
{{- if .Values.customLabels }}
{{ toYaml .Values.customLabels | indent 4 }}
{{- end }}
{{- end }}
+40 -2
View File
@@ -51,7 +51,7 @@ rbac:
# isClusterService specifies whether chart should be deployed as cluster-service or normal k8s app.
isClusterService: true
# Optional priority class to be used for the coredns pods
# Optional priority class to be used for the coredns pods. Used for autoscaler if autoscaler.priorityClassName not set.
priorityClassName: ""
# Default zone is what Kubernetes recommends:
@@ -147,5 +147,43 @@ extraSecrets: []
# - name: some-fancy-secret
# mountPath: /etc/wherever
# Custom labels to apply to Deployment, Pod, Service, ServiceMonitor
# Custom labels to apply to Deployment, Pod, Service, ServiceMonitor. Including autoscaler if enabled.
customLabels: {}
## Configue a cluster-proportional-autoscaler for coredns
# See https://github.com/kubernetes-incubator/cluster-proportional-autoscaler
autoscaler:
# Enabled the cluster-proportional-autoscaler
enabled: false
# Number of cores in the cluster per coredns replica
coresPerReplica: 256
# Number of nodes in the cluster per coredns replica
nodesPerReplica: 16
image:
repository: k8s.gcr.io/cluster-proportional-autoscaler-amd64
tag: "1.7.1"
pullPolicy: IfNotPresent
# Optional priority class to be used for the autoscaler pods. priorityClassName used if not set.
priorityClassName: ""
# expects input structure as per specification https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.11/#affinity-v1-core
affinity: {}
# Node labels for pod assignment
# Ref: https://kubernetes.io/docs/user-guide/node-selection/
nodeSelector: {}
# expects input structure as per specification https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.11/#toleration-v1-core
tolerations: []
# resources for autoscaler pod
resources:
requests:
cpu: "20m"
memory: "10Mi"
limits:
cpu: "20m"
memory: "10Mi"