StatefulSet

StatefulSet 是为了解决有状态服务的问题(对应 Deployments 和 ReplicaSets 是为无状态服务而设计),其应用场景包括

  • 稳定的持久化存储,即 Pod 重新调度后还是能访问到相同的持久化数据,基于 PVC 来实现

  • 稳定的网络标志,即 Pod 重新调度后其 PodName 和 HostName 不变,基于 Headless Service(即没有 Cluster IP 的 Service)来实现

  • 有序部署,有序扩展,即 Pod 是有顺序的,在部署或者扩展的时候要依据定义的顺序依次依序进行(即从 0 到 N-1,在下一个 Pod 运行之前所有之前的 Pod 必须都是 Running 和 Ready 状态),基于 init containers 来实现

  • 有序收缩,有序删除(即从 N-1 到 0)

从上面的应用场景可以发现,StatefulSet 由以下几个部分组成:

  • 用于定义网络标志(DNS domain)的 Headless Service

  • 用于创建 PersistentVolumes 的 volumeClaimTemplates

  • 定义具体应用的 StatefulSet

StatefulSet 中每个 Pod 的 DNS 格式为 statefulSetName-{0..N-1}.serviceName.namespace.svc.cluster.local,其中

  • serviceName 为 Headless Service 的名字

  • 0..N-1 为 Pod 所在的序号,从 0 开始到 N-1

  • statefulSetName 为 StatefulSet 的名字

  • namespace 为服务所在的 namespace,Headless Service 和 StatefulSet 必须在相同的 namespace

  • .cluster.local 为 Cluster Domain

API 版本对照表

Kubernetes 版本Deployment 版本

v1.5-v1.6

extensions/v1beta1

v1.7-v1.15

apps/v1beta1

v1.8-v1.15

apps/v1beta2

v1.9+

apps/v1

简单示例

以一个简单的 nginx 服务 web.yaml 为例:

apiVersion: v1
kind: Service
metadata:
  name: nginx
  labels:
    app: nginx
spec:
  ports:
  - port: 80
    name: web
  clusterIP: None
  selector:
    app: nginx
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: web
spec:
  serviceName: "nginx"
  replicas: 2
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: k8s.gcr.io/nginx-slim:0.8
        ports:
        - containerPort: 80
          name: web
        volumeMounts:
        - name: www
          mountPath: /usr/share/nginx/html
  volumeClaimTemplates:
  - metadata:
      name: www
    spec:
      accessModes: ["ReadWriteOnce"]
      resources:
        requests:
          storage: 1Gi
$ kubectl create -f web.yaml
service "nginx" created
statefulset "web" created

# 查看创建的 headless service 和 statefulset
$ kubectl get service nginx
NAME      CLUSTER-IP   EXTERNAL-IP   PORT(S)   AGE
nginx     None         <none>        80/TCP    1m
$ kubectl get statefulset web
NAME      DESIRED   CURRENT   AGE
web       2         2         2m

# 根据 volumeClaimTemplates 自动创建 PVC(在 GCE 中会自动创建 kubernetes.io/gce-pd 类型的 volume)
$ kubectl get pvc
NAME        STATUS    VOLUME                                     CAPACITY   ACCESSMODES   AGE
www-web-0   Bound     pvc-d064a004-d8d4-11e6-b521-42010a800002   1Gi        RWO           16s
www-web-1   Bound     pvc-d06a3946-d8d4-11e6-b521-42010a800002   1Gi        RWO           16s

# 查看创建的 Pod,他们都是有序的
$ kubectl get pods -l app=nginx
NAME      READY     STATUS    RESTARTS   AGE
web-0     1/1       Running   0          5m
web-1     1/1       Running   0          4m

# 使用 nslookup 查看这些 Pod 的 DNS
$ kubectl run -i --tty --image busybox dns-test --restart=Never --rm /bin/sh
/ # nslookup web-0.nginx
Server:    10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local

Name:      web-0.nginx
Address 1: 10.244.2.10
/ # nslookup web-1.nginx
Server:    10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local

Name:      web-1.nginx
Address 1: 10.244.3.12
/ # nslookup web-0.nginx.default.svc.cluster.local
Server:    10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local

Name:      web-0.nginx.default.svc.cluster.local
Address 1: 10.244.2.10

还可以进行其他的操作

# 扩容
$ kubectl scale statefulset web --replicas=5

# 缩容
$ kubectl patch statefulset web -p '{"spec":{"replicas":3}}'

# 镜像更新(目前还不支持直接更新 image,需要 patch 来间接实现)
$ kubectl patch statefulset web --type='json' -p='[{"op":"replace","path":"/spec/template/spec/containers/0/image","value":"gcr.io/google_containers/nginx-slim:0.7"}]'

# 删除 StatefulSet 和 Headless Service
$ kubectl delete statefulset web
$ kubectl delete service nginx

# StatefulSet 删除后 PVC 还会保留着,数据不再使用的话也需要删除
$ kubectl delete pvc www-web-0 www-web-1

更新 StatefulSet

v1.7 + 支持 StatefulSet 的自动更新,通过 spec.updateStrategy 设置更新策略。目前支持两种策略

  • OnDelete:当 .spec.template 更新时,并不立即删除旧的 Pod,而是等待用户手动删除这些旧 Pod 后自动创建新 Pod。这是默认的更新策略,兼容 v1.6 版本的行为

  • RollingUpdate:当 .spec.template 更新时,自动删除旧的 Pod 并创建新 Pod 替换。在更新时,这些 Pod 是按逆序的方式进行,依次删除、创建并等待 Pod 变成 Ready 状态才进行下一个 Pod 的更新。

Partitions

RollingUpdate 还支持 Partitions,通过 .spec.updateStrategy.rollingUpdate.partition 来设置。当 partition 设置后,只有序号大于或等于 partition 的 Pod 会在 .spec.template 更新的时候滚动更新,而其余的 Pod 则保持不变(即便是删除后也是用以前的版本重新创建)。

# 设置 partition 为 3
$ kubectl patch statefulset web -p '{"spec":{"updateStrategy":{"type":"RollingUpdate","rollingUpdate":{"partition":3}}}}'
statefulset "web" patched

# 更新 StatefulSet
$ kubectl patch statefulset web --type='json' -p='[{"op":"replace","path":"/spec/template/spec/containers/0/image","value":"gcr.io/google_containers/nginx-slim:0.7"}]'
statefulset "web" patched

# 验证更新
$ kubectl delete po web-2
pod "web-2" deleted
$ kubectl get po -lapp=nginx -w
NAME      READY     STATUS              RESTARTS   AGE
web-0     1/1       Running             0          4m
web-1     1/1       Running             0          4m
web-2     0/1       ContainerCreating   0          11s
web-2     1/1       Running             0          18s

Pod 管理策略

v1.7 + 可以通过 .spec.podManagementPolicy 设置 Pod 管理策略,支持两种方式

  • OrderedReady:默认的策略,按照 Pod 的次序依次创建每个 Pod 并等待 Ready 之后才创建后面的 Pod

  • Parallel:并行创建或删除 Pod(不等待前面的 Pod Ready 就开始创建所有的 Pod)

Parallel 示例

---
apiVersion: v1
kind: Service
metadata:
  name: nginx
  labels:
    app: nginx
spec:
  ports:
  - port: 80
    name: web
  clusterIP: None
  selector:
    app: nginx
---
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
  name: web
spec:
  serviceName: "nginx"
  podManagementPolicy: "Parallel"
  replicas: 2
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: gcr.io/google_containers/nginx-slim:0.8
        ports:
        - containerPort: 80
          name: web
        volumeMounts:
        - name: www
          mountPath: /usr/share/nginx/html
  volumeClaimTemplates:
  - metadata:
      name: www
    spec:
      accessModes: ["ReadWriteOnce"]
      resources:
        requests:
          storage: 1Gi

可以看到,所有 Pod 是并行创建的

$ kubectl create -f webp.yaml
service "nginx" created
statefulset "web" created

$ kubectl get po -lapp=nginx -w
NAME      READY     STATUS              RESTARTS  AGE
web-0     0/1       Pending             0         0s
web-0     0/1       Pending             0         0s
web-1     0/1       Pending             0         0s
web-1     0/1       Pending             0         0s
web-0     0/1       ContainerCreating   0         0s
web-1     0/1       ContainerCreating   0         0s
web-0     1/1       Running             0         10s
web-1     1/1       Running             0         10s

zookeeper

另外一个更能说明 StatefulSet 强大功能的示例为 zookeeper.yaml

---
apiVersion: v1
kind: Service
metadata:
  name: zk-headless
  labels:
    app: zk-headless
spec:
  ports:
  - port: 2888
    name: server
  - port: 3888
    name: leader-election
  clusterIP: None
  selector:
    app: zk
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: zk-config
data:
  ensemble: "zk-0;zk-1;zk-2"
  jvm.heap: "2G"
  tick: "2000"
  init: "10"
  sync: "5"
  client.cnxns: "60"
  snap.retain: "3"
  purge.interval: "1"
---
apiVersion: policy/v1beta1
kind: PodDisruptionBudget
metadata:
  name: zk-budget
spec:
  selector:
    matchLabels:
      app: zk
  minAvailable: 2
---
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
  name: zk
spec:
  serviceName: zk-headless
  replicas: 3
  template:
    metadata:
      labels:
        app: zk
      annotations:
        pod.alpha.kubernetes.io/initialized: "true"
        scheduler.alpha.kubernetes.io/affinity: >
            {
              "podAntiAffinity": {
                "requiredDuringSchedulingRequiredDuringExecution": [{
                  "labelSelector": {
                    "matchExpressions": [{
                      "key": "app",
                      "operator": "In",
                      "values": ["zk-headless"]
                    }]
                  },
                  "topologyKey": "kubernetes.io/hostname"
                }]
              }
            }
    spec:
      containers:
      - name: k8szk
        imagePullPolicy: Always
        image: gcr.io/google_samples/k8szk:v1
        resources:
          requests:
            memory: "4Gi"
            cpu: "1"
        ports:
        - containerPort: 2181
          name: client
        - containerPort: 2888
          name: server
        - containerPort: 3888
          name: leader-election
        env:
        - name : ZK_ENSEMBLE
          valueFrom:
            configMapKeyRef:
              name: zk-config
              key: ensemble
        - name : ZK_HEAP_SIZE
          valueFrom:
            configMapKeyRef:
                name: zk-config
                key: jvm.heap
        - name : ZK_TICK_TIME
          valueFrom:
            configMapKeyRef:
                name: zk-config
                key: tick
        - name : ZK_INIT_LIMIT
          valueFrom:
            configMapKeyRef:
                name: zk-config
                key: init
        - name : ZK_SYNC_LIMIT
          valueFrom:
            configMapKeyRef:
                name: zk-config
                key: tick
        - name : ZK_MAX_CLIENT_CNXNS
          valueFrom:
            configMapKeyRef:
                name: zk-config
                key: client.cnxns
        - name: ZK_SNAP_RETAIN_COUNT
          valueFrom:
            configMapKeyRef:
                name: zk-config
                key: snap.retain
        - name: ZK_PURGE_INTERVAL
          valueFrom:
            configMapKeyRef:
                name: zk-config
                key: purge.interval
        - name: ZK_CLIENT_PORT
          value: "2181"
        - name: ZK_SERVER_PORT
          value: "2888"
        - name: ZK_ELECTION_PORT
          value: "3888"
        command:
        - sh
        - -c
        - zkGenConfig.sh && zkServer.sh start-foreground
        readinessProbe:
          exec:
            command:
            - "zkOk.sh"
          initialDelaySeconds: 15
          timeoutSeconds: 5
        livenessProbe:
          exec:
            command:
            - "zkOk.sh"
          initialDelaySeconds: 15
          timeoutSeconds: 5
        volumeMounts:
        - name: datadir
          mountPath: /var/lib/zookeeper
      securityContext:
        runAsUser: 1000
        fsGroup: 1000
  volumeClaimTemplates:
  - metadata:
      name: datadir
      annotations:
        volume.alpha.kubernetes.io/storage-class: anything
    spec:
      accessModes: ["ReadWriteOnce"]
      resources:
        requests:
          storage: 20Gi
kubectl create -f zookeeper.yaml

详细的使用说明见 zookeeper stateful application

StatefulSet 注意事项

  1. 推荐在 Kubernetes v1.9 或以后的版本中使用

  2. 所有 Pod 的 Volume 必须使用 PersistentVolume 或者是管理员事先创建好

  3. 为了保证数据安全,删除 StatefulSet 时不会删除 Volume

  4. StatefulSet 需要一个 Headless Service 来定义 DNS domain,需要在 StatefulSet 之前创建好

最后更新于