重构郑重声明！！！

[cuisongliu]

在项目的发展过程中，最初采用了Service和Endpoint的方法来解决外部服务的特性问题，以实现对集群外部服务的访问。然而，逐渐发现这种方法无法满足一些有状态服务的需求。其中一个主要原因是Service和Endpoint的定义只允许相同的端口，不能使用不同的端口作为后端地址。为了解决这个问题，项目决定废弃这一设计，重新构思一个新的Operator。

新的Operator设计具备以下特点和能力：

1. 重复使用lvscare的一部分功能，直接透传IPVS的能力，实现后端地址的自由组合。
2. 每个ClusterIP都具备独立的探活功能，支持TCP、UDP、GRPC、HTTP等不同的协议。
3. 支持自动域名管理的能力。

最终目标是允许接入有状态服务，而无需每个主机都有相同的端口设置。这一能力非常重要，未来可以借助这个功能实现外部中间件内部的负载均衡，无需担心云上不同集群之间的访问问题。

[cuisongliu]

1. 给对应的redis创建service 配置

apiVersion: v1
kind: Service
metadata:
  annotations:
    monitor.kubeblocks.io/agamotto: "false"
    monitor.kubeblocks.io/scrape: "false"
  labels:
    app.kubernetes.io/component: redis
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    app.kubernetes.io/name: redis
    apps.kubeblocks.io/component-name: redis
  name: redis-redis-headless-p0
  namespace: redis-cloud
spec:
  ports:
  - name: redis
    port: 6379
    protocol: TCP
    targetPort: redis
  selector:
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    apps.kubeblocks.io/component-name: redis
    statefulset.kubernetes.io/pod-name: redis-redis-0
  sessionAffinity: None
  type: NodePort
status:
  loadBalancer: {}
---

apiVersion: v1
kind: Service
metadata:
  annotations:
    monitor.kubeblocks.io/agamotto: "false"
    monitor.kubeblocks.io/scrape: "false"
  labels:
    app.kubernetes.io/component: redis
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    app.kubernetes.io/name: redis
    apps.kubeblocks.io/component-name: redis
  name: redis-redis-headless-p1
  namespace: redis-cloud
spec:
  ports:
  - name: redis
    port: 6379
    protocol: TCP
    targetPort: redis
  selector:
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    apps.kubeblocks.io/component-name: redis
    statefulset.kubernetes.io/pod-name: redis-redis-1
  sessionAffinity: None
  type: NodePort
status:
  loadBalancer: {}
---
apiVersion: v1
kind: Service
metadata:
  annotations:
    monitor.kubeblocks.io/agamotto: "false"
    monitor.kubeblocks.io/scrape: "false"
  labels:
    app.kubernetes.io/component: redis-sentinel
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    app.kubernetes.io/name: redis
    apps.kubeblocks.io/component-name: redis-sentinel
  name: redis-redis-sentinel-headless-p0
  namespace: redis-cloud
spec:
  ports:
  - name: redis-sentinel
    port: 26379
    protocol: TCP
    targetPort: redis-sentinel
  selector:
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    apps.kubeblocks.io/component-name: redis-sentinel
    statefulset.kubernetes.io/pod-name: redis-redis-sentinel-0
  type: NodePort
status:
  loadBalancer: {}
---
apiVersion: v1
kind: Service
metadata:
  annotations:
    monitor.kubeblocks.io/agamotto: "false"
    monitor.kubeblocks.io/scrape: "false"
  labels:
    app.kubernetes.io/component: redis-sentinel
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    app.kubernetes.io/name: redis
    apps.kubeblocks.io/component-name: redis-sentinel
  name: redis-redis-sentinel-headless-p1
  namespace: redis-cloud
spec:
  ports:
  - name: redis-sentinel
    port: 26379
    protocol: TCP
    targetPort: redis-sentinel
  selector:
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    apps.kubeblocks.io/component-name: redis-sentinel
    statefulset.kubernetes.io/pod-name: redis-redis-sentinel-1
  type: NodePort
status:
  loadBalancer: {}
---
apiVersion: v1
kind: Service
metadata:
  annotations:
    monitor.kubeblocks.io/agamotto: "false"
    monitor.kubeblocks.io/scrape: "false"
  labels:
    app.kubernetes.io/component: redis-sentinel
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    app.kubernetes.io/name: redis
    apps.kubeblocks.io/component-name: redis-sentinel
  name: redis-redis-sentinel-headless-p2
  namespace: redis-cloud
spec:
  ports:
  - name: redis-sentinel
    port: 26379
    protocol: TCP
    targetPort: redis-sentinel
  selector:
    app.kubernetes.io/instance: redis
    app.kubernetes.io/managed-by: kubeblocks
    apps.kubeblocks.io/component-name: redis-sentinel
    statefulset.kubernetes.io/pod-name: redis-redis-sentinel-2
  type: NodePort
status:
  loadBalancer: {}

[cuisongliu]

2. 调kube-proxy 的配置 防止自动删除ipvs规则

excludeCIDRs:
      - 10.103.97.2/32
      - 30.96.0.0/16

3. 新增对应的ipvs规则

ipvsadm -Ln


ipvsadm -D -t 30.96.0.1:6379
ipvsadm -A -t 30.96.0.1:6379 -s rr
ipvsadm -a -t 30.96.0.1:6379 -r 172.18.2.70:30362  -m


ipvsadm -D -t 30.96.0.2:6379
ipvsadm -A -t 30.96.0.2:6379 -s rr
ipvsadm -a -t 30.96.0.2:6379 -r 172.18.2.70:32065  -m



ipvsadm -D -t 30.96.0.3:26379
ipvsadm -A -t 30.96.0.3:26379 -s rr
ipvsadm -a -t 30.96.0.3:26379 -r 172.18.2.70:31012  -m


ipvsadm -D -t 30.96.0.4:26379
ipvsadm -A -t 30.96.0.4:26379 -s rr
ipvsadm -a -t 30.96.0.4:26379 -r 172.18.2.70:30655  -m




ipvsadm -D -t 30.96.0.5:26379
ipvsadm -A -t 30.96.0.5:26379 -s rr
ipvsadm -a -t 30.96.0.5:26379 -r 172.18.2.70:30300  -m




30.96.0.1 redis-redis-0.redis-redis-headless.redis-cloud.svc.cluster.local 
30.96.0.2 redis-redis-1.redis-redis-headless.redis-cloud.svc.cluster.local
30.96.0.3 redis-redis-sentinel-0.redis-redis-sentinel-headless.redis-cloud.svc.cluster.local 
30.96.0.4 redis-redis-sentinel-1.redis-redis-sentinel-headless.redis-cloud.svc.cluster.local 
30.96.0.5 redis-redis-sentinel-2.redis-redis-sentinel-headless.redis-cloud.svc.cluster.local


client:  redis-redis-sentinel-0.redis-redis-sentinel-headless.redis-cloud.svc.cluster.local:26379,redis-redis-sentinel-1.redis-redis-sentinel-headless.redis-cloud.svc.cluster.local:26379,redis-redis-sentinel-2.redis-redis-sentinel-headless.redis-cloud.svc.cluster.local:26379

[cuisongliu]

手动创建规则如何被pod之间访问？

[cuisongliu]

https://zhangguanzhang.github.io/2021/09/28/ipvs-svc/#/配置-keepalived

[cuisongliu]

https://github.com/labring/lvscare