部署集群过程中遇到的坑：

1  node(s) had taints that the pod didn't tolerate.

创新互联致力于网站建设,网站制作设计,营销网页按需设计网站,外贸网站制作,企业网站建设,小程序设计,网站SEO优化,网站设计制作案例丰富,是成都做网站公司和建站公司,欢迎咨询。

集群中有个节点被污染了，查看节点信息，发现其中一个节点没有起来，导致只能启动2个pod

2 PV和PVC上的坑

• 一个zk.yaml中需要创建3个zookeeper，根据zookeeper集群的特性，三个zookeeper需要三个不同的数据目录（如果再添加日志目录也是一样），集群才会正常起来

• 创建PV的时候需要创建三个不同的PV

• 在一个zk.yaml中，PVC需要关联到后端不同的三个PVC，如果按照传统的创建PVC，并在zk.yaml中指定PVC的名字，只能指定一个名字，这样是办不到关联到三个不同的PV，于是需要volumeClaimTemplates（PVC模板），模板accessModes需要ReadWriteOnce,ReadWriteMany会关联失败。

  volumeClaimTemplates:
  - metadata:
      name: datadir
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 1Gi

• 如果是用动态pvc供给，也是一样需要创建PVC模板

3 zookeeper提供给外部访问

zookeeper由于是使用statefulSet部署，使用到了无头服务，如果提供给外部访问呢？办法就是增加一个专门为外部访问的service。

具体部署

1 在192.168.0.11上创建安装nfs，并配置启动nfs

2 在每个node节点上下载nfs

3 在matser节点创建zk使用的PV

apiVersion: v1
kind: PersistentVolume
metadata:
  name: zk-data1
spec:
  capacity:
    storage: 1Gi
  accessModes:
    - ReadWriteOnce
  nfs:
    server: 192.168.0.11
    path: /data/zk/data1
---
apiVersion: v1
kind: PersistentVolume
metadata:
  name: zk-data2
spec:
  capacity:
    storage: 1Gi
  accessModes:
    - ReadWriteOnce
  nfs:
    server: 192.168.0.11
    path: /data/zk/data2
---
apiVersion: v1
kind: PersistentVolume
metadata:
  name: zk-data3
spec:
  capacity:
    storage: 1Gi
  accessModes:
    - ReadWriteOnce
  nfs:
    server: 192.168.0.11
    path: /data/zk/data3

4 在master节点上创建zookeeper的创建pod的yaml文件

参考官方文档：https://kubernetes.io/docs/tutorials/stateful-application/zookeeper/https://kubernetes.io/docs/tutorials/stateful-application/zookeeper/

官方文档中的镜像地址为国外地址，我这里网上找了一个可用的地址，跟官方镜像一样

[root@localhost zk]# cat zk.yaml 
apiVersion: v1
kind: Service
metadata:
  name: zk-hs
  labels:
    app: zk
spec:
  ports:
  - port: 2888
    name: server
  - port: 3888
    name: leader-election
  clusterIP: None
  selector:
    app: zk
---
apiVersion: v1
kind: Service
metadata:
  name: zk-cs
  labels:
    app: zk
spec:
  type: NodePort
  ports:
  - port: 2181
    targetPort: 2181
    name: client
    nodePort: 2181
  selector:
    app: zk
---
apiVersion: policy/v1beta1
kind: PodDisruptionBudget
metadata:
  name: zk-pdb
spec:
  selector:
    matchLabels:
      app: zk
  maxUnavailable: 1
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: zok
spec:
  serviceName: zk-hs
  replicas: 3
  selector:
    matchLabels:
      app: zk
  template:
    metadata:
      labels:
        app: zk
    spec:
      affinity:
        podAntiAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            - labelSelector:
                matchExpressions:
                  - key: "app"
                    operator: In
                    values:
                    - zk
              topologyKey: "kubernetes.io/hostname"
      containers:
      - name: kubernetes-zookeeper
        imagePullPolicy: Always
        image: leolee32/kubernetes-library:kubernetes-zookeeper1.0-3.4.10
        resources:
          requests:
            memory: "1Gi"
            cpu: "0.5"
        ports:
        - containerPort: 2181
          name: client
        - containerPort: 2888
          name: server
        - containerPort: 3888
          name: leader-election
        command:
        - sh
        - -c
        - "start-zookeeper \
          --servers=3 \
          --data_dir=/var/lib/zookeeper/data \
          --data_log_dir=/var/lib/zookeeper/data/log \
          --conf_dir=/opt/zookeeper/conf \
          --client_port=2181 \
          --election_port=3888 \
          --server_port=2888 \
          --tick_time=2000 \
          --init_limit=10 \
          --sync_limit=5 \
          --heap=512M \
          --max_client_cnxns=60 \
          --snap_retain_count=3 \
          --purge_interval=12 \
          --max_session_timeout=40000 \
          --min_session_timeout=4000 \
          --log_level=INFO"
        readinessProbe:
          exec:
            command:
            - sh
            - -c
            - "zookeeper-ready 2181"
          initialDelaySeconds: 10
          timeoutSeconds: 5
        livenessProbe:
          exec:
            command:
            - sh
            - -c
            - "zookeeper-ready 2181"
          initialDelaySeconds: 10
          timeoutSeconds: 5
        volumeMounts:
        - name: datadir
          mountPath: /var/lib/zookeeper
  volumeClaimTemplates:
  - metadata:
      name: datadir
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 1Gi

5 查看集群状态

for i in 0 1 2; do kubectl exec zok-$i zkServer.sh status; done

6 外部访问zookeeper集群（使用本机电脑访问）

新闻名称：kubernetes部署有状态应用之zookeeper集群
分享地址：http://ybzwz.com/article/pedsce.html