1、需求
使用K8S部署server1，server2两个服务，server2调用server1的/test1接口，需支持负载均衡
2、实现两个服务用于测试
pyserver1.py
from flask import Flask, jsonifyapp = Flask(__name__)@app.route('/test1', methods=['GET'])
def test1():return jsonify({"message": "This is Server1!"})if __name__ == '__main__':app.run(host='0.0.0.0', port=5001)
pyserver2.py
import requests
from flask import Flask, jsonifyapp = Flask(__name__)SERVER1_URL = "http://server1-service.default.svc.cluster.local/test1"@app.route('/test2', methods=['GET'])
def test2():try:response = requests.get(SERVER1_URL)response.raise_for_status()return jsonify(response.json()), response.status_codeexcept requests.RequestException as e:return jsonify({"error": str(e)}), 500if __name__ == '__main__':app.run(host='0.0.0.0', port=5002)
注意API地址http://server1-service.default.svc.cluster.local:80/test1
与后续的server1-delpoyment-service.yaml的service配置对应：
3、制作docker镜像
- export DOCKER_BUILDKIT=0 #解决某些版本如docker-ce:18.05构建镜像缓慢问题
- 基础镜像使用阿里云私有仓库拉的一个python镜像，先从阿里云容器镜像服务web页面搜索到这个镜像再docker pull，拉取前需要docker login一下，账号密码同等web页面的，需要配置该账号的权限，都登录了那就不需配置/etc/docker/daemon.json的insecure-registries，registry-mirrors。
- server2镜像的制作只需要“1”改为“2”即可
FROM alibaba-cloud-linux-3-registry.cn-hangzhou.cr.aliyuncs.com/alinux3/python:3.11.1WORKDIR /appRUN pip install --no-cache-dir requests flask -i http://mirrors.aliyun.com/pypi/simple --trusted-host mirrors.aliyun.comCOPY ./pyserver1.py .CMD ["python", "pyserver1.py"]
执行构建：
docker build -f Dockerfile-server1 -t server1 .
docker build -f Dockerfile-server2 -t server2 .
4、编写服务部署的yaml文件
server1-deployment-service.yaml
apiVersion: apps/v1
kind: Deployment
metadata:name: server1
spec:replicas: 2selector:matchLabels:app: label-server1template:metadata:labels:app: label-server1spec:containers:- name: server1-instanceimage: server1:latestimagePullPolicy: Neverports:- containerPort: 5001
---
apiVersion: v1
kind: Service
metadata:name: server1-servicenamespace: default
spec:selector:app: label-server1ports:- protocol: TCPport: 80targetPort: 5001type: LoadBalancer
server2-deployment-service.yaml
apiVersion: apps/v1
kind: Deployment
metadata:name: server2
spec:replicas: 2selector:matchLabels:app: label-server2template:metadata:labels:app: label-server2spec:containers:- name: server2-instanceimage: server2:latestimagePullPolicy: Neverports:- containerPort: 5002
---
apiVersion: v1
kind: Service
metadata:name: server2-servicenamespace: default
spec:selector:app: label-server2ports:- protocol: TCPport: 80targetPort: 5002type: LoadBalancer
去除master的不可调度污点让server1,server2可以部署一个实例到master：kubectl taint nodes 128master node-role.kubernetes.io/master:NoSchedule-
部署应用并暴露服务：
kubectl apply -f server1-deployment-service.yaml
kubectl apply -f server2-deployment-service.yaml
5、接口调用测试
请求server2的/test2接口，该接口调用了server1的/test1接口
看看负载均衡效果，如果不需要复杂的负载均衡策略，基于K8S实现微服务还是可以的。
附：K8S calico网络问题
在测试接口调用主要出现了两个网络问题。
我的K8S集群初始化命令
kubeadm init --apiserver-advertise-address=192.168.72.128 --image-repository registry.aliyuncs.com/google_containers
calico 直接apply的官方找的原文件。如果kubeadm init时指定了--pod-network-cidr,这时caclio的yamlCALICO_IPV4POOL_CIDR配置项也对应起来即可。我的K8S之前安装好后，过了一段时间没用，coredns组件异常或没有ready了。这时curl localhost:30309/test2会报错请求http://server1-service.default.svc.cluster.local:80/test1时域名解析不了。
[root@128aworker ~]# kubectl get all -A -o wide
NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
kube-system pod/calico-kube-controllers-577f77cb5c-94cxn 1/1 Running 0 2m21s 10.244.35.1 127amaster <none> <none>
kube-system pod/calico-node-89pdk 0/1 CrashLoopBackOff 4 2m21s 192.168.72.128 128aworker <none> <none>
kube-system pod/calico-node-bt4qx 1/1 Running 0 2m21s 192.168.72.127 127amaster <none> <none>
kube-system pod/coredns-6d56c8448f-kfcx9 1/1 Running 0 6m26s 10.244.55.64 128aworker <none> <none>
kube-system pod/coredns-6d56c8448f-zlbcb 1/1 Running 0 6m26s 10.244.55.65 128aworker <none> <none>
kube-system pod/etcd-128aworker 1/1 Running 0 6m40s 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-apiserver-128aworker 1/1 Running 0 6m40s 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-controller-manager-128aworker 1/1 Running 0 6m40s 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-proxy-dk4mj 1/1 Running 0 6m26s 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-proxy-n8cnw 1/1 Running 0 6m14s 192.168.72.127 127amaster <none> <none>
kube-system pod/kube-scheduler-128aworker 1/1 Running 0 6m40s 192.168.72.128 128aworker <none> <none>NAMESPACE NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
default service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 6m43s <none>
kube-system service/kube-dns ClusterIP 10.96.0.10 <none> 53/UDP,53/TCP,9153/TCP 6m42s k8s-app=kube-dns
master worker都kubeadm reset -f重置后重新建立集群coredns OK了，但caclio还没有ready
[root@128aworker ~]# kubectl get all -A -o wide
NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
kube-system pod/calico-kube-controllers-577f77cb5c-94cxn 1/1 Running 0 8m29s 10.244.35.1 127amaster <none> <none>
kube-system pod/calico-node-89pdk 0/1 Running 6 8m29s 192.168.72.128 128aworker <none> <none>
kube-system pod/calico-node-bt4qx 0/1 Running 0 8m29s 192.168.72.127 127amaster <none> <none>
kube-system pod/coredns-6d56c8448f-kfcx9 1/1 Running 0 12m 10.244.55.64 128aworker <none> <none>
kube-system pod/coredns-6d56c8448f-zlbcb 1/1 Running 0 12m 10.244.55.65 128aworker <none> <none>
kube-system pod/etcd-128aworker 1/1 Running 0 12m 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-apiserver-128aworker 1/1 Running 0 12m 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-controller-manager-128aworker 1/1 Running 0 12m 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-proxy-dk4mj 1/1 Running 0 12m 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-proxy-n8cnw 1/1 Running 0 12m 192.168.72.127 127amaster <none> <none>
kube-system pod/kube-scheduler-128aworker 1/1 Running 0 12m 192.168.72.128 128aworker <none> <none>
再次测试发现http://server1-service.default.svc.cluster.local:80/test1网络不通，logs/describe pod都找不到明显的错误解决思路，其实这是docker自定义host_elastic bridge网络污染了K8S的网络，或许高版本K8S/calico已经解决了该问题
calico issue 相关问题还比较多。。。
[root@128aworker ~]# docker network ls
NETWORK ID NAME DRIVER SCOPE
b11900aaa7fb bridge bridge local
d2685ac3c5a5 home_elastic bridge local
d00ae1cde2a2 host host local
47c26a5dffe2 none null local
[root@128aworker ~]# docker network rm d26
删除所有节点docker自定义网络并kubectl delete -f calico.yaml再重新apply即可
[root@128aworker home]# kubectl get all -A -o wide
NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
kube-system pod/calico-kube-controllers-577f77cb5c-bgrxd 1/1 Running 0 49s 10.244.35.1 127amaster <none> <none>
kube-system pod/calico-node-fbcxz 1/1 Running 0 49s 192.168.72.128 128aworker <none> <none>
kube-system pod/calico-node-jq2kc 1/1 Running 0 49s 192.168.72.127 127amaster <none> <none>
kube-system pod/coredns-6d56c8448f-qxfhg 1/1 Running 0 2m10s 10.244.55.66 128aworker <none> <none>
kube-system pod/coredns-6d56c8448f-zvrj8 1/1 Running 0 2m10s 10.244.55.65 128aworker <none> <none>
kube-system pod/etcd-128aworker 1/1 Running 0 2m24s 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-apiserver-128aworker 1/1 Running 0 2m24s 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-controller-manager-128aworker 1/1 Running 0 2m24s 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-proxy-2w87z 1/1 Running 0 106s 192.168.72.127 127amaster <none> <none>
kube-system pod/kube-proxy-tdf7k 1/1 Running 0 2m10s 192.168.72.128 128aworker <none> <none>
kube-system pod/kube-scheduler-128aworker 1/1 Running 0 2m24s 192.168.72.128 128aworker <none> <none>NAMESPACE NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
default service/kubernetes ClusterIP 10.1.0.1 <none> 443/TCP 2m27s <none>
kube-system service/kube-dns ClusterIP 10.1.0.10 <none> 53/UDP,53/TCP,9153/TCP 2m25s k8s-app=kube-dns