[cloud native | kubernetes] in depth RC, RS, daemonset, statefulset (VII)

​ thorough RC、RS、DaemonSet、StatefulSet

One 、RC、RS

RC （ReplicationController ） The main function is to ensure that the number of copies of the container application is always kept at the user-defined number of copies . That is, if a container exits abnormally , Will automatically create a new Pod To replace ; And if the extra containers are recovered automatically

Kubernetes Official recommended use RS（ReplicaSet ） replace RC （ReplicationController ） Deployment ,RS Follow RC There is no essential difference , It's just that the names are different , also RS Support for aggregate selector

RC controller

apiVersion: v1
kind: ReplicationController
metadata:
  name: frontend
spec:
  replicas: 3
  selector:
    app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: php-redis
        image: lansonlinux/myapp:v1
        env:
        - name: GET_HOSTS_FROM
          value: dns
          name: zhangsan
          value: "123"
        ports:
        - containerPort: 80

Copy

RS controller

apiVersion: apps/v1
kind: ReplicaSet
metadata:
  name: frontend
spec:
  replicas: 3
  selector:
    matchLabels:
      tier: frontend
  template:
    metadata:
      labels:
        tier: frontend
    spec:
      containers:
      - name: myapp
        image: lansonlinux/myapp:v1
        env:
        - name: GET_HOSTS_FROM
          value: dns
        ports:
        - containerPort: 80

Copy

Two 、DaemonSet

DaemonSet The controller ensures that all （ Or part of it ） All nodes run a specified Pod copy .

• Whenever a node is added to the cluster , designated Pod A copy will also be added to the node
• When a node is removed from the cluster ,Pod It's recycled
• Delete one DaemonSet You can clean up all the created Pod

DaemonSet The typical usage scenarios are ：

• Run the cluster's storage daemon on each node , for example glusterd、ceph
• Run the log collection daemons on each node , for example fluentd、logstash
• Run monitoring daemons on each node , for example Prometheus Node Exporter、Sysdig Agent、collectd、Dynatrace OneAgent、APPDynamics Agent、Datadog agent、New Relic agent、Ganglia gmond、Instana Agent etc.

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: logging
  labels:
    app: logging
spec:
  selector:
    matchLabels:
      name: logging
  template:
    metadata:
      labels:
        name: logging
    spec:
      containers:
      - name: logging
        image: nginx
        resources:
          limits:
            memory: 200Mi
          requests:
            cpu: 100m
            memory: 200Mi
      tolerations:  # Set tolerance master The stain of 
      - key: node-role.kubernetes.io/master
        effect: NoSchedule

Copy

See the effect

kubectl get pod -l name=logging -o wide

Copy

3、 ... and 、StatefulSet

Stateful replica set ;Deployment Such as stateless application deployment （stateless）

• StatefulSet Use scenarios ; For applications with the following requirements ,StatefulSet just the thing ：
  • Stable 、 The only Internet logo （dnsname）
    • StatefulSet Through its associated headless service for each pod Provide DNS Parse entry . If there is no head service DNS The entry is : "$(service name).$(namespace).svc.cluster.local", that pod The parsing entry is "$(pod name).$(service name).$(namespace).svc.cluster.local", Every pod name It's the only one .
  • The stability of the 、 Persistent storage ;【 Every Pod Always correspond to their respective storage paths （PersistantVolumeClaimTemplate）】
  • Orderly 、 Elegant deployment and scaling .【 Add copies in order 、 Reduce copies , And perform cleanup when reducing copies 】
  • Orderly 、 Automatic rollover update .【 Rolling updates are performed automatically in sequence 】
• Limit
  • Given Pod Must be stored by PersistentVolume The driver is based on the requested storage class To provide , Or by the administrator in advance .
  • Delete or shrink StatefulSet and Can't Delete its associated storage volume . This is to ensure data security , It's usually better than automatically clearing StatefulSet All the related resources are more valuable .
  • StatefulSet Currently, headless services are needed to be responsible for Pod Network identity of . You need to be responsible for creating this service .
  • When the delete StatefulSets when ,StatefulSet No termination is provided Pod Guarantee . In order to achieve StatefulSet Medium Pod Can be terminated in an orderly and dignified manner , You can delete StatefulSet Zoom in to 0.
  • By default Pod Management strategy (OrderedReady) When using Scroll to update , It is possible to enter a damaged state that requires human intervention to repair .

If an application does not need a stable network identity , Or you don't need to deploy in order 、 Delete 、 Add a copy of the , We should consider using Deployment This kind of Statelessness （stateless） The controller

apiVersion: v1 kind: Service # Define a load balancing network metadata: name: stateful-tomcat labels: app: stateful-tomcat spec: ports: - port: 8123 name: web targetPort: 8080 clusterIP: None #NodePort： Any machine +NodePort All have access to ,ClusterIP： This can be used in the cluster ip、service Domain name can access ,clusterIP: None; Do not assign clusters ip.headless; The headless service . Stable domain name selector: app: stateful-tomcat --- apiVersion: apps/v1 kind: StatefulSet # controller . metadata: name: stateful-tomcat spec: selector: matchLabels: app: stateful-tomcat # has to match .spec.template.metadata.labels serviceName: "stateful-tomcat" # Pay attention here , There must be a service The name is this replicas: 3 # by default is 1 template: metadata: labels: app: stateful-tomcat # has to match .spec.selector.matchLabels spec: terminationGracePeriodSeconds: 10 containers: - name: tomcat image: tomcat:7 ports: - containerPort: 8080 name: web # Observe the effect . Delete one , Name after restart ,ip It's all the same . Ensure the State # details kubectl explain StatefulSet.spec podManagementPolicy： OrderedReady（ In order ）、Parallel（ Concurrent ） serviceName -required- Set the service name , You can use the domain name to access pod 了 . pod-specific-string.serviceName.default.svc.cluster.local # test kubectl run -i --tty --image busybox dns-test --restart=Never --rm /bin/sh ping stateful-tomcat-0.stateful-tomcat # We do not add storage volumes here . If any kubectl get pvc -l app=stateful-tomcat We can see that even if Pod Delete it and then pull it up , The volume is the same .

​