k8s中部署redis集群

上一篇中，我们在k8s中部署了一个单节点的redis。这一节就继续升级难度，部署一个redis集群。

redis集群和redis单节点差别很大，而且相比于在服务器上部署redis集群，在k8s环境下的部署又麻烦了一些。

原因在于服务器的ip的固定不变的，但是容器环境下pod的ip每次重启都会改变。

redis cluster维持集群又需要确人ip。所以需要我们处理pod重启的情况。

yaml文件

configmap文件

apiVersion: v1
kind: ConfigMap
metadata:
  name: redis-config
  namespace: redis-test
data:
  fix-ip.sh: |
    #!/bin/sh
    CLUSTER_CONFIG="/data/data/nodes.conf"
    if [ -f ${CLUSTER_CONFIG} ]; then
      if [ -z "${POD_IP}" ]; then
        echo "miss Pod IP address!"
        exit 1
      fi
      echo "Updating my IP to ${POD_IP} in ${CLUSTER_CONFIG}"
      sed -i.bak -e '/myself/ s/[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}/'${POD_IP}'/' ${CLUSTER_CONFIG}
    fi
    exec "$@"
  redis.conf: |
    dir /data/data
    requirepass 123456
    logfile /data/data/redis.log
    cluster-enabled yes
    cluster-config-file /data/data/nodes.conf
    protected-mode no
    daemonize no
    pidfile /var/run/redis.pid
    port 6379
    bind 0.0.0.0
    timeout 3600
    tcp-keepalive 1
    loglevel verbose
    databases 16
    save 900 1
    save 300 10
    save 60 10000
    rdbcompression yes
    rdbchecksum yes
    dbfilename dump.rdb
    appendonly yes
    appendfilename "appendonly.aof"
    appendfsync everysec
    篇幅原因，很多配置省略了

注意这个fix-ip.sh脚本，他的作用是在pod重启时，将自己之前生成的nodes.conf文件中myself的ip改成新的。

其他活着的pod能够自动更新新的ip，只有自己需要处理下。

statefulset文件

apiVersion: v1
kind: Service
metadata:
  namespace: redis-test
  name: redis-cluster
spec:
  clusterIP: None
  ports:
    - port: 6379
      targetPort: 6379
      name: client
    - port: 16379
      targetPort: 16379
      name: gossport
  selector:
    app: redis-cluster
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  namespace: redis-test
  name: redis-cluster
spec:
  serviceName: redis-cluster
  replicas: 6
  selector:
    matchLabels:
      app: redis-cluster
  template:
    metadata:
      labels:
        app: redis-cluster
    spec:
      containers:
        - name: redis
          image: redis:5.0.5-alpine
          ports:
            - containerPort: 6379
              name: client
            - containerPort: 16379
              name: gossport
          command: ["/data/conf/fix-ip.sh", "redis-server", "/data/conf/redis.conf"]
          env:
            - name: POD_IP
              valueFrom:
                fieldRef:
                  fieldPath: status.podIP
          volumeMounts:
            - name: config
              mountPath: /data/conf
              readOnly: false
            - name: data
              mountPath: /data/data
              readOnly: false
      volumes:
        - name: config
          configMap:
            name: redis-config
            defaultMode: 0755
  volumeClaimTemplates:
    - metadata:
        name: data
        annotations:
          volume.beta.kubernetes.io/storage-class: "nfs-client"
      spec:
        accessModes:
          - ReadWriteMany
        resources:
          requests:
            storage: 2Gi

我们使用statefulset的volumeClaimTemplates，为每一个pod生成其对应的pvc和pv

service

apiVersion: v1
kind: Service
metadata:
  namespace: redis-test
  name: redis-cluster
spec:
  type: ClusterIP
  ports:
    - port: 6379
      targetPort: 6379
      name: client
  selector:
    app: redis-cluster

好了。

分别执行。全部启动

[root@paas-m-k8s-master-1 ~]# kc -n redis-test get pod,sts,svc
NAME                  READY   STATUS    RESTARTS   AGE
pod/redis-cluster-0   1/1     Running   0          3h3m
pod/redis-cluster-1   1/1     Running   0          14m
pod/redis-cluster-2   1/1     Running   0          3h3m
pod/redis-cluster-3   1/1     Running   0          3h3m
pod/redis-cluster-4   1/1     Running   0          3h3m
pod/redis-cluster-5   1/1     Running   0          3h2m


NAME                             READY   AGE
statefulset.apps/redis-cluster   6/6     3h3m


NAME                    TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)              AGE
service/redis-cluster   ClusterIP   None         <none>        6379/TCP,16379/TCP   33m

组装集群

我们在服务器上组件集群会用以下命令

redis-cli -a 123456 --cluster create  ip:port ip2:port ip3:port ip4:port ip5:port ip6:port  --cluster-replicas 1

我们又可以用以下命令获取所有pod的ip

kubectl -n redis-test get pods -l app=redis-cluster -o jsonpath='{range.items[*]}{.status.podIP}:6379 '

所以连起来就是

kubectl -n redis-test exec -it redis-cluster-0 -- redis-cli -a 123456 --cluster create $(kubectl -n redis-test get pods -l app=redis-cluster -o jsonpath='{range.items[*]}{.status.podIP}:6379 ') --cluster-replicas 1

好了，搭建完成。

去pod里连接一下看看

用service连接一下看看

[root@paas-m-k8s-master-1 redis]# kc -n yys exec -it myredis-sample-0 -- redis-cli -c -h redis-cluster.redis-test -a 123456
Warning: Using a password with '-a' or '-u' option on the command line interface may not be safe.
redis-cluster.redis-test:6379> cluster info
cluster_state:ok
cluster_slots_assigned:16384
cluster_slots_ok:16384
...

注意，现在我们的redis在集群外是不能连接的，就算把service设置为nodeport模式也不行。因为现在redis配置的ip是k8s集群内的ip。

在外面不能被识别。

目前我们的策略是生产环境就应该是生产环境集群内使用，本来也不应该被外界访问。

这样确实在需要排查问题时会比较麻烦，我们单独开发了一个web页面用来供用户被授权后查询使用。

另外有个取巧的方法是，部署一个predixy进行redis的代理。

测试pod重启

先写一个数据aaa，看到这个数据存到了100.105.152.12这个pod

[root@paas-m-k8s-master-1 ~]# kc -n yys exec -it myredis-sample-0 -- redis-cli -c -h redis-cluster.redis-test -a 123456
Warning: Using a password with '-a' or '-u' option on the command line interface may not be safe.
redis-cluster.redis-test:6379> set aaa 2023
-> Redirected to slot [10439] located at 100.105.152.12:6379
OK
100.105.152.12:6379>

看到这个pod是redis-cluster-1

[root@paas-m-k8s-master-1 ~]# kc -n redis-test get pod -owide
NAME              READY   STATUS    RESTARTS   AGE    IP                NODE                NOMINATED NODE   READINESS GATES
redis-cluster-0   1/1     Running   0          168m   100.66.220.252    paas-m-k8s-node-1   <none>           <none>
redis-cluster-1   1/1     Running   0          168m   100.105.152.12    paas-m-k8s-node-4   <none>           <none>
redis-cluster-2   1/1     Running   0          168m   100.115.23.183    paas-m-k8s-node-5   <none>           <none>
redis-cluster-3   1/1     Running   0          168m   100.66.54.124     paas-m-k8s-node-3   <none>           <none>
redis-cluster-4   1/1     Running   0          168m   100.108.161.168   paas-m-k8s-node-2   <none>           <none>
redis-cluster-5   1/1     Running   0          167m   100.111.149.223   paas-m-k8s-node-6   <none>           <none>

删除redis-cluster-1 。新启动的pod的ip变为100.105.152.25

[root@paas-m-k8s-master-1 ~]# kc -n redis-test delete pod redis-cluster-1
pod "redis-cluster-1" deleted
[root@paas-m-k8s-master-1 ~]# kc -n redis-test get pod -owide
NAME              READY   STATUS    RESTARTS   AGE    IP                NODE                NOMINATED NODE   READINESS GATES
redis-cluster-0   1/1     Running   0          169m   100.66.220.252    paas-m-k8s-node-1   <none>           <none>
redis-cluster-1   1/1     Running   0          4s     100.105.152.25    paas-m-k8s-node-4   <none>           <none>
redis-cluster-2   1/1     Running   0          168m   100.115.23.183    paas-m-k8s-node-5   <none>           <none>
redis-cluster-3   1/1     Running   0          168m   100.66.54.124     paas-m-k8s-node-3   <none>           <none>
redis-cluster-4   1/1     Running   0          168m   100.108.161.168   paas-m-k8s-node-2   <none>           <none>
redis-cluster-5   1/1     Running   0          168m   100.111.149.223   paas-m-k8s-node-6   <none>           <none>

查看集群的数据ok

[root@paas-m-k8s-master-1 ~]# kc -n myhome exec -it myredis-sample-0 -- redis-cli -c -h redis-cluster.redis-test -a 123456
Warning: Using a password with '-a' or '-u' option on the command line interface may not be safe.
redis-cluster.redis-test:6379> get aaa
-> Redirected to slot [10439] located at 100.105.152.25:6379
"2023"

说明持久化和fix-ip是ok的

为什么会使用fix-ip这种蹩脚的方式？

其实一开始的时候，测试使用podName.svcName.nameSpace这种statefulset特有的访问方式，因为statefulset这种headless service方式来说，其podName.svcName.nameSpace命名是固定不变的，所以以podName.svcName.nameSpace次方式来构建redis cluster简直太合适了。

kubectl -n redis-test exec -it redis-cluster-0 -- redis-cli -a 123456 --cluster create redis-cluster-0.redis-cluster.redis-test:6379 redis-cluster-1.redis-cluster.redis-test:6379 redis-cluster-2.redis-cluster.redis-test:6379 redis-cluster-3.redis-cluster.redis-test:6379 redis-cluster-4.redis-cluster.redis-test:6379 redis-cluster-5.redis-cluster.redis-test:6379 --cluster-replicas 1

不过非常可惜，redis有检验，这种方式不是合法的ip

另外ix-ip这种方案有一个风险，那就是如果多个redis实例同时宕机重启。也就是cluster集群已经fail，那么整个集群无法自我修复。需要人工干预。