zookeeper和k8s_k8s部署有状态服务zookeeper示例

我们先不考虑配置文件的前提下：

apiVersion: apps/v1

kind: StatefulSet #####固定hostname，有状态的服务使用这个 statefalset有个问题，就是如果那个pod不是running状态，这个主机名是无法解析的，这样就构成了一个死循环，我sed替换主机名的时候由于pod还不是running状态，她只能获取自己的主机名。无法获取别人的主机名，所以在zookeeper中换成了换成了ip

metadata:

name: zookeeper

spec:

serviceName: zookeeper ####所以生成的3个pod的名字叫zookeeper-0,zookeeper-1,zookeeper-2

replicas: 3

revisionHistoryLimit: 10

selector: ##statefulset必须有的

matchLabels:

app: zookeeper

template:

metadata:

labels:

app: zookeeper

spec:

volumes:

- name: volume-logs

hostPath:

path: /var/log/zookeeper

containers:

- name: zookeeper

image: harbor.test.com/middleware/zookeeper:3.4.10

imagePullPolicy: IfNotPresent

livenessProbe:

tcpSocket:

port: 2181

initialDelaySeconds: 30

timeoutSeconds: 3

periodSeconds: 5

successThreshold: 1

failureThreshold: 2

ports:

- containerPort: 2181

protocol: TCP

- containerPort: 2888

protocol: TCP

- containerPort: 3888

protocol: TCP

env:

- name: SERVICE_NAME

value: "zookeeper"

- name: MY_POD_NAME #声明k8s自带的变量，这样在pod创建之后，在其中可以直接echo ${MY_POD_NAME}得到hostname

valueFrom:

fieldRef:

fieldPath: metadata.name

volumeMounts:

- name: volume-logs

mountPath: /var/log/zookeeper

nodeSelector:

zookeeper: enable

---

apiVersion: v1

kind: Service

metadata:

name: zookeeper #我的cluster名字为这个，在任意一个生成的pod中可以ping zookeeper，相当于zookeeper为生成的3个pod的cluster_name,会发现每次ping出的地址不一定相同，nslookup zookeeper得到的是3个pod的pod ip，共3条记录。

spec:

ports:

- port: 2181

selector:

app: zookeeper

clusterIP: None #此句必须加上

[root@host5 src]# kubectl get pod --all-namespaces -o wide

NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES

default zookeeper-0 1/1 Running 0 12m 192.168.55.69 host3

default zookeeper-1 1/1 Running 0 12m 192.168.31.93 host4

default zookeeper-2 1/1 Running 0 12m 192.168.55.70 host3

bash-4.3# nslookup zookeeper

nslookup: can't resolve '(null)': Name does not resolve

Name: zookeeper

Address 1: 192.168.55.70 zo