Helm部署Ingress使用DaemonSet+Taint/Tolerations+NodeSelector

kubernetes集群中需要在指定的几个节点上只部署Nginx Ingress Controller实例,不会跑其他业务容器。

环境说明

教程基于有k84集群,并安装好helm部署环境。强烈推荐使用helm发布您的代码!

执行helm version出现如下证明环境已经就绪。

  1. [root@oneinstack ~]# helm version
  2. Client: &version.Version{SemVer:"v2.13.1", GitCommit:"618447cbf203d147601b4b9bd7f8c37a5d39fbb4", GitTreeState:"clean"}
  3. Server: &version.Version{SemVer:"v2.13.1", GitCommit:"618447cbf203d147601b4b9bd7f8c37a5d39fbb4", GitTreeState:"clean"}

helm下载ingress chart

搜索ingress charts

  1. helm search ingress

下载nginx ingress

  1. helm fetch stable/nginx-ingress

fetch之后得到nginx-ingress-1.4.0.tgz

修改ingress helm chart

解压nginx-ingress-1.4.0.tgz

代码语言:javascript
复制
tar xzf nginx-ingress-1.4.0.tgz
注意: 解压过程出现implausibly old time stamp 1970-01-01 08:00:00可忽略
修改values.yaml,下面是我的修改好的:
## nginx configuration
## Ref: https://github.com/kubernetes/ingress/blob/master/controllers/nginx/configuration.md
##
controller:
  name: controller
  image:
    repository: quay.io/kubernetes-ingress-controller/nginx-ingress-controller  # 建议将镜像拖到自己私有参考,修改私有仓库地址
    tag: "0.24.1"
    pullPolicy: IfNotPresent
    # www-data -> uid 33
    runAsUser: 33
  config: {}
  # Will add custom header to Nginx https://github.com/kubernetes/ingress-nginx/tree/master/docs/examples/customization/custom-headers
  headers: {}
  # Required for use with CNI based kubernetes installations (such as ones set up by kubeadm),
  # since CNI and hostport don't mix yet. Can be deprecated once https://github.com/kubernetes/kubernetes/issues/23920
  # is merged
  hostNetwork: true   # 80 443 暴露到宿主机
  # Optionally change this to ClusterFirstWithHostNet in case you have 'hostNetwork: true'.
  # By default, while using host network, name resolution uses the host's DNS. If you wish nginx-controller
  # to keep resolving names inside the k8s network, use ClusterFirstWithHostNet.
  dnsPolicy: ClusterFirst
  ## Use host ports 80 and 443
  daemonset:
    useHostPort: false
    hostPorts:
      http: 80
      https: 443
      ## healthz endpoint
      stats: 18080
  ## Required only if defaultBackend.enabled = false
  ## Must be <namespace>/<service_name>
  ##
  defaultBackendService: ""
  ## Election ID to use for status update
  ##
  electionID: ingress-controller-leader
  ## Name of the ingress class to route through this controller
  ##
  ingressClass: nginx # 后续ingress.yaml annotations指定kubernetes.io/ingress.class: nginx
  # labels to add to the pod container metadata
  podLabels: {}
  #  key: value
  ## Allows customization of the external service
  ## the ingress will be bound to via DNS
  publishService:
    enabled: false
    ## Allows overriding of the publish service to bind to
    ## Must be <namespace>/<service_name>
    ##
    pathOverride: ""
  ## Limit the scope of the controller
  ##
  scope:
    enabled: false
    namespace: ""   # defaults to .Release.Namespace
  ## Additional command line arguments to pass to nginx-ingress-controller
  ## E.g. to specify the default SSL certificate you can use
  ## extraArgs:
  ##   default-ssl-certificate: "<namespace>/<secret_name>"
  extraArgs: {}
  ## Additional environment variables to set
  extraEnvs: []
  # extraEnvs:
  #   - name: FOO
  #     valueFrom:
  #       secretKeyRef:
  #         key: FOO
  #         name: secret-resource
  ## DaemonSet or Deployment
  ##
  kind: DaemonSet   #DaemonSet模式
  # The update strategy to apply to the Deployment or DaemonSet
  ##
  updateStrategy: {}
  #  rollingUpdate:
  #    maxUnavailable: 1
  #  type: RollingUpdate
  # minReadySeconds to avoid killing pods before we are ready
  ##
  minReadySeconds: 0
  ## Node tolerations for server scheduling to nodes with taints
  ## Ref: https://kubernetes.io/docs/concepts/configuration/assign-pod-node/
  ##
  tolerations:    #在节点上打污点,此处是容忍的key value effect
    - key: "nginx-ingress"
      operator: "Equal"
      value: "true"
      effect: "NoSchedule"
  affinity: {}
  ## Node labels for controller pod assignment
  ## Ref: https://kubernetes.io/docs/user-guide/node-selection/
  ##
  nodeSelector:
    nginx-ingress: "true"    #使用节点标签选择器,访问在所有节点运行ingress
  ## Liveness and readiness probe values
  ## Ref: https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle/#container-probes
  ##
  livenessProbe:
    failureThreshold: 3
    initialDelaySeconds: 10
    periodSeconds: 10
    successThreshold: 1
    timeoutSeconds: 1
    port: 10254
  readinessProbe:
    failureThreshold: 3
    initialDelaySeconds: 10
    periodSeconds: 10
    successThreshold: 1
    timeoutSeconds: 1
    port: 10254
  ## Annotations to be added to controller pods
  ##
  podAnnotations:    # 支持prometheus抓取数据
    prometheus.io/scrape: "true"
    prometheus.io/port: "10254"
  replicaCount: 1
  minAvailable: 1
  resources: {}
  #  limits:
  #    cpu: 100m
  #    memory: 64Mi
  #  requests:
  #    cpu: 100m
  #    memory: 64Mi
  autoscaling:
    enabled: false
    minReplicas: 1
    maxReplicas: 11
    targetCPUUtilizationPercentage: 50
    targetMemoryUtilizationPercentage: 50
  ## Override NGINX template
  customTemplate:
    configMapName: ""
    configMapKey: ""
  service:
    annotations: {}
    labels: {}
    clusterIP: ""
    ## List of IP addresses at which the controller services are available
    ## Ref: https://kubernetes.io/docs/user-guide/services/#external-ips
    ##
    externalIPs: []
    loadBalancerIP: ""
    loadBalancerSourceRanges: []
    enableHttp: true
    enableHttps: true
    ## Set external traffic policy to: "Local" to preserve source IP on
    ## providers supporting it
    ## Ref: https://kubernetes.io/docs/tutorials/services/source-ip/#source-ip-for-services-with-typeloadbalancer
    externalTrafficPolicy: ""
    healthCheckNodePort: 0
    targetPorts:
      http: http
      https: https
    type: ClusterIP
    # type: NodePort
    # nodePorts:
    #   http: 32080
    #   https: 32443
    nodePorts:
      http: ""
      https: ""
  extraContainers: []
  ## Additional containers to be added to the controller pod.
  ## See https://github.com/lemonldap-ng-controller/lemonldap-ng-controller as example.
  #  - name: my-sidecar
  #    image: nginx:latest
  #  - name: lemonldap-ng-controller
  #    image: lemonldapng/lemonldap-ng-controller:0.2.0
  #    args:
  #      - /lemonldap-ng-controller
  #      - --alsologtostderr
  #      - --configmap=$(POD_NAMESPACE)/lemonldap-ng-configuration
  #    env:
  #      - name: POD_NAME
  #        valueFrom:
  #          fieldRef:
  #            fieldPath: metadata.name
  #      - name: POD_NAMESPACE
  #        valueFrom:
  #          fieldRef:
  #            fieldPath: metadata.namespace
  #    volumeMounts:
  #    - name: copy-portal-skins
  #      mountPath: /srv/var/lib/lemonldap-ng/portal/skins
  extraVolumeMounts: []
  ## Additional volumeMounts to the controller main container.
  #  - name: copy-portal-skins
  #   mountPath: /var/lib/lemonldap-ng/portal/skins
  extraVolumes: []
  ## Additional volumes to the controller pod.
  #  - name: copy-portal-skins
  #    emptyDir: {}
  extraInitContainers: []
  ## Containers, which are run before the app containers are started.
  # - name: init-myservice
  #   image: busybox
  #   command: ['sh', '-c', 'until nslookup myservice; do echo waiting for myservice; sleep 2; done;']
  stats:
    enabled: true
    service:
      annotations: {}
      clusterIP: ""
      ## List of IP addresses at which the stats service is available
      ## Ref: https://kubernetes.io/docs/user-guide/services/#external-ips
      ##
      externalIPs: []
      loadBalancerIP: ""
      loadBalancerSourceRanges: []
      servicePort: 18080
      type: ClusterIP
  ## If controller.stats.enabled = true and controller.metrics.enabled = true, Prometheus metrics will be exported
  ##
  metrics:
    enabled: true
    service:
      annotations:
        prometheus.io/scrape: "true"
        prometheus.io/port: "10254"
      clusterIP: ""
      ## List of IP addresses at which the stats-exporter service is available
      ## Ref: https://kubernetes.io/docs/user-guide/services/#external-ips
      ##
      externalIPs: []
      loadBalancerIP: ""
      loadBalancerSourceRanges: []
      servicePort: 9913
      type: ClusterIP
    serviceMonitor:
      enabled: false
      additionalLabels: {}
      namespace: ""
  lifecycle: {}
  priorityClassName: ""
## Rollback limit
##
revisionHistoryLimit: 10
## Default 404 backend
##
defaultBackend:
  ## If false, controller.defaultBackendService must be provided
  ##
  enabled: true
  name: default-backend
  image:
    repository: registry.cn-hangzhou.aliyuncs.com/google_containers/defaultbackend
    tag: "1.4"
    pullPolicy: IfNotPresent
  extraArgs: {}
  port: 8080
  ## Node tolerations for server scheduling to nodes with taints
  ## Ref: https://kubernetes.io/docs/concepts/configuration/assign-pod-node/
  ##
  tolerations: []
  #  - key: "key"
  #    operator: "Equal|Exists"
  #    value: "value"
  #    effect: "NoSchedule|PreferNoSchedule|NoExecute(1.6 only)"
  affinity: {}
  # labels to add to the pod container metadata
  podLabels: {}
  #  key: value
  ## Node labels for default backend pod assignment
  ## Ref: https://kubernetes.io/docs/user-guide/node-selection/
  ##
  nodeSelector: {}
  ## Annotations to be added to default backend pods
  ##
  podAnnotations: {}
  replicaCount: 1
  minAvailable: 1
  resources: {}
  # limits:
  #   cpu: 10m
  #   memory: 20Mi
  # requests:
  #   cpu: 10m
  #   memory: 20Mi
  service:
    annotations: {}
    clusterIP: ""
    ## List of IP addresses at which the default backend service is available
    ## Ref: https://kubernetes.io/docs/user-guide/services/#external-ips
    ##
    externalIPs: []
    loadBalancerIP: ""
    loadBalancerSourceRanges: []
    servicePort: 80
    type: ClusterIP
  priorityClassName: ""
## Enable RBAC as per https://github.com/kubernetes/ingress/tree/master/examples/rbac/nginx and https://github.com/kubernetes/ingress/issues/266
rbac:
  create: true
# If true, create & use Pod Security Policy resources
# https://kubernetes.io/docs/concepts/policy/pod-security-policy/
podSecurityPolicy:
  enabled: false
serviceAccount:
  create: true
  name:
## Optional array of imagePullSecrets containing private registry credentials
## Ref: https://kubernetes.io/docs/tasks/configure-pod-container/pull-image-private-registry/
imagePullSecrets: []
# - name: secretName
# TCP service key:value pairs
# Ref: https://github.com/kubernetes/contrib/tree/master/ingress/controllers/nginx/examples/tcp
##
tcp: {}
#  8080: "default/example-tcp-svc:9000"
# UDP service key:value pairs
# Ref: https://github.com/kubernetes/contrib/tree/master/ingress/controllers/nginx/examples/udp
##
udp: {}
#  53: "kube-system/kube-dns:53"

NODE打标签,设置污点

打标签

  1. kubectl label nodes 10.0.10.7 nginx-ingress=true
  2. kubectl label nodes 10.0.10.8 nginx-ingress=true
  3. kubectl label nodes 10.0.10.9 nginx-ingress=true

设置污点

  1. kubectl taint nodes 10.0.10.7 nginx-ingress=true:NoSchedule
  2. kubectl taint nodes 10.0.10.8 nginx-ingress=true:NoSchedule
  3. kubectl taint nodes 10.0.10.9 nginx-ingress=true:NoSchedule

安装nginx-ingress

  1. helm upgrade nginx-ingress ./nginx-ingress --install --namespace nginx-ingress --dry-run  # 测试运行
  2. helm upgrade nginx-ingress ./nginx-ingress --install --namespace nginx-ingress

注意:

  1. 需要先在NODE节点打污点、标签
  2. helm名字和命名空间请使用nginx-ingress, 和直接用yaml文件(ingress-nginx)有区别。否则DaemonSet、pod名字比较奇怪