准备环境
先把 google 的vm 跑起来…
gateman@MoreFine-S500:~/projects/coding/k8s-s/service-case/cloud-user$ kubectl get nodes
NAME STATUS ROLES AGE VERSION
k8s-master Ready control-plane,master 124d v1.23.6
k8s-node0 Ready <none> 124d v1.23.6
k8s-node1 NotReady <none> 124d v1.23.6
k8s-node3 Ready <none> 104d v1.23.6
当前的环境是干净的
gateman@MoreFine-S500:~/projects/coding/k8s-s/service-case/cloud-user$ kubectl get all -o wide
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 81d <none>
准备1个 micro-service 和 1个clusterIP
大概框架如上图:
组件
micro service
我会先构建1个 微服务 bq-api-service 并部署到集群, 设成4个pods
clusterIP
我会构建1个 ClusterIP service: clusterip-bq-api-service 作为上面微服务的loadbanlance 和 revserse proxy
测试pod dns-test
因为没有nodePort 和 ingress, 只能用测试POD 去尝试通过 clusterip service 链接 bq-api-serivce
部署bq-api-service
deployment-bq-api-service.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
labels: # label of this deployment
app: bq-api-service # custom defined
author: nvd11
name: deployment-bq-api-service # name of this deployment
namespace: default
spec:
replicas: 4 # desired replica count, Please note that the replica Pods in a Deployment are typically distributed across multiple nodes.
revisionHistoryLimit: 10 # The number of old ReplicaSets to retain to allow rollback
selector: # label of the Pod that the Deployment is managing,, it's mandatory, without it , we will get this error
# error: error validating data: ValidationError(Deployment.spec.selector): missing required field "matchLabels" in io.k8s.apimachinery.pkg.apis.meta.v1.LabelSelector ..
matchLabels:
app: bq-api-service
strategy: # Strategy of upodate
type: RollingUpdate # RollingUpdate or Recreate
rollingUpdate:
maxSurge: 25% # The maximum number of Pods that can be created over the desired number of Pods during the update
maxUnavailable: 25% # The maximum number of Pods that can be unavailable during the update
template: # Pod template
metadata:
labels:
app: bq-api-service # label of the Pod that the Deployment is managing. must match the selector, otherwise, will get the error Invalid value: map[string]string{"app":"bq-api-xxx"}: `selector` does not match template `labels`
spec:
containers:
- image: europe-west2-docker.pkg.dev/jason-hsbc/my-docker-repo/bq-api-service:1.1.7 # image of the container
imagePullPolicy: IfNotPresent
name: bq-api-service-container
env: # set env varaibles
- name: APP_ENVIRONMENT
value: prod
restartPolicy: Always # Restart policy for all containers within the Pod
terminationGracePeriodSeconds: 10 # The period of time in seconds given to the Pod to terminate gracefully
---
apiVersion: v1
kind: Service
metadata:
name: clusterip-bq-api-service
spec:
selector:
app: bq-api-service # for the pods that have the label app: bq-api-service
ports:
- protocol: TCP
port: 8080
targetPort: 8080
type: ClusterIP
执行命令:
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/test$ kubectl apply -f bq-api-service.yml
deployment.apps/deployment-bq-api-service created
service/clusterip-bq-api-service unchanged
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/test$ kubectl get ep -o wide
NAME ENDPOINTS AGE
clusterip-bq-api-service 10.244.1.70:8080,10.244.1.71:8080,10.244.2.141:8080 + 1 more... 36s
kubernetes 192.168.0.3:6443 83d
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/test$ kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
deployment-bq-api-service-6f6ffc7866-2mn5n 1/1 Running 0 27s 10.244.2.141 k8s-node0 <none> <none>
deployment-bq-api-service-6f6ffc7866-74bcq 1/1 Running 0 27s 10.244.1.70 k8s-node1 <none> <none>
deployment-bq-api-service-6f6ffc7866-l2xzb 1/1 Running 0 27s 10.244.1.71 k8s-node1 <none> <none>
deployment-bq-api-service-6f6ffc7866-lwxt7 1/1 Running 0 27s 10.244.3.82 k8s-node3 <none> <none>
看起来没什么问题
测试cluster ip 的连接
先进入dns-test
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/test$ kubectl run dns-test --image=odise/busybox-curl --restart=Never -- /bin/sh -c "while true; do echo hello docker; sleep 1; done"
pod/dns-test created
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/test$ kubectl exec -it dns-test -- /bin/sh
/ #
测试调用clusterip service
/ # curl clusterip-bq-api-service:8080/actuator/info
{"app":"Sales API","version":"1.1.7","hostname":"deployment-bq-api-service-6f6ffc7866-2mn5n","description":"This is a simple Spring Boot application to demonstrate the use of BigQuery in GCP."}/ #
/ # curl clusterip-bq-api-service:8080/actuator/info
/ # curl clusterip-bq-api-service:8080/actuator/info
{"app":"Sales API","version":"1.1.7","hostname":"deployment-bq-api-service-6f6ffc7866-l2xzb","description":"This is a simple Spring Boot application to demonstrate the use of BigQuery in GCP."}/ #
看来 reverse proxy 和 loadbalance 都生效~
安装helm
下载:
https://github.com/helm/helm/releases
配置~/.bashrc 就好
export PATH=$PATH:/home/gateman/devtools/helm
测试version
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/test$ helm version
version.BuildInfo{
Version:"v3.15.2", GitCommit:"1a500d5625419a524fdae4b33de351cc4f58ec35", GitTreeState:"clean", GoVersion:"go1.22.4"}
通过helm 安装 ingress-nginx controller
把 ingress-ngnix 的repo 添加到helm
helm repo add ingress-nginx https://kubernetes.github.io/ingress-nginx
查看repo list
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/test$ helm repo list
NAME URL
ingress-nginx https://kubernetes.github.io/ingress-nginx
helm 下载 ingress-nginx package
helm pull <package-name>
helm pull ingress-nginx/ingress-nginx
这时, 会下载1个tgz 文件 ingress-nginx-4.10.1.tgz
解压:
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/ingress-nginx/helm/ingress-nginx$ ls -l
total 132
drwxrwxr-x 2 gateman gateman 4096 6月 25 01:26 changelog
-rw-r--r-- 1 gateman gateman 752 4月 26 22:02 Chart.yaml
drwxrwxr-x 2 gateman gateman 4096 6月 25 01:26 ci
-rw-r--r-- 1 gateman gateman 177 4月 26 22:02 OWNERS
-rw-r--r-- 1 gateman gateman 49134 4月 26 22:02 README.md
-rw-r--r-- 1 gateman gateman 11358 4月 26 22:02 README.md.gotmpl
drwxrwxr-x 3 gateman gateman 4096 6月 25 01:26 templates
drwxrwxr-x 2 gateman gateman 4096 6月 25 01:26 tests
-rw-r--r-- 1 gateman gateman 45136 6月 25 02:02 values.yaml
其中values.yaml 就是最终要的配置文件
修改values.yaml
有些教程让我们修改 image repo 的 address 指向阿里云加速, 由于我的server在google cloud 就忽略这一步了。
-
controller.kind -> 由 Deployment 改成 DaemonSet
这样做的优点是
a. 能够保证每个Node节点上都运行一个 ingress-nginx Pod实例,实现全集群访问入口。(实际上我们会用label 只让ingress-nginx 部署在1台node上)
b. 节点故障转移时,新的Node上会自动启动Pod,保证入口始终可用。 -
controller.hostnetwork -> 由false 改成 true
这是DaemonSet的默认网络模式。每个Pod将直接共享Node的网络栈和资源,简化了网络配置。
但同时也会带来上面提到的一些隔离性和安全性问题。 -
controller.dnsPolic-> 由 ClusterFirst 改成 ClusterFirstWithHostNet
ClusterFirst表示使用集群提供的DNS服务进行解析,这也是默认策略。
但是DaemonSet使用hostNetwork模式后,Pod会直接使用宿主机的网络命名空间。
此时设置dnsPolicy=ClusterFirstWithHostNet表示:
首先使用集群的DNS服务来解析域名。
如果集群DNS无法响应,则解析请求会转发到宿主机的DNS服务上寻求响应。
这样可以实现两个目的:
使用集群的DNS服务提高可配置性和维护性。
同时如果集群DNSdown掉,Pod也可以fallback到宿主机本地DNS上,保证关键服务名能被解析到。
这相比直接使用宿主机DNS或仅使用集群DNS更加灵活。
因此对于使用hostNetwork的DaemonSet,dnsPolicy设置为ClusterFirstWithHostNet可以充分利用集群和宿主机的DNS功能,提高服务可用性。 -
controller.service.type -> 由 Loadbalancer 改成 ClusterIP
对于用Helm部署ingress-nginx,使用ClusterIP作为service类型通常是更好的选择。
LoadBalancer类型会要求底层基础设施如云平台自动为ingress分配公网IP地址,但这可能会产生额外的网络开支。
考虑到大多数K8S集群运行在内网环境,使用ClusterIP就可以满足需求。后续还可以配置支持内网访问的解决方案,如NodePort或External IPs。
一般来说,你可以直接在values.yaml里修改service.type为ClusterIP:
在部署 ingress-nginx controller 之前(我们已经把它改成1个DaemonSet), 为它创建1个单独的namespace
kubectl create ns ingress-nginx
我们之后会单独地把 ingress-ingress 部署在这个 同名字的namespace 上面。
问题来了, 在命名空间A的 ingress controller 可以用与命名空间B的ingress 实例吗?
答案是yes:
关于Ingress controller namespace与Ingress资源namespace的关系:
1. Ingress controller一般会部署在单独的namespace下,比如ingress-namespace。
2. Ingress资源可以定义在任何namespace中。
3. Kubernetes会检测所有namespace下是否有Ingress Controller的服务运行。
如果有,则Ingress资源对应的namespace会自动重定向到Ingress controller所在的namespace。
Ingress controller通过监视Apiserver,可以查询到集群内所有namespace下定义的Ingress资源。
也就是说:
1. Ingress controller和Ingress资源可以部署在不同的namespace。
2. Kubernetes会自动帮助Ingress资源找到部署在哪个namespace的Ingress controller。
3. Ingress controller 通过Apiserver可以查到所有namespace下的Ingress定义。
为所在node添加label
ingress-controller 作为1个类网关程序, 理论上只在其中1个node 部署就够了
由于我们修改了value.yaml, 把这个controll的类型改成了DaemonSet 所以, 理论上每个node 都会部署1个pod, 是有资源浪费的。
而实际上, value.yaml 有定义了1个规则- node selector, 只有具有某个label ingress=true 的node 才会部署
node selector 的定义:
controller.nodeSelector:
nodeSelector: # nodeSelector of DaemonSet we updated
kubernetes.io/os: linux
ingress: "true" # use String instead of bool
所以我们只需要为1台node 添加这个label 就好
但是, 如果我们只给master node添加label是不work的, 因为master node 配置了污点, 会bypass 大部分 deployment/DaemonSet 的部署
这次我们只给k8s-node0 这个node 添加label
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/ingress-nginx/helm/ingress-nginx$ kubectl label no k8s-node0 ingress=true
node/k8s-node0 labeled
查看label
gateman@MoreFine-S500:~/tmp/ingress-nginx$ kubectl get nodes --show-labels=true
NAME STATUS ROLES AGE VERSION LABELS
k8s-master Ready control-plane,master 128d v1.23.6 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,ingress=true,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node-role.kubernetes.io/master=,node.kubernetes.io/exclude-from-external-load-balancers=
k8s-node0 Ready <none> 128d v1.23.6 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,ingress=true,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node0,kubernetes.io/os=linux
k8s-node1 Ready <none> 128d v1.23.6 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node1,kubernetes.io/os=linux
k8s-node3 Ready <none> 108d v1.23.6 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node3,kubernetes.io/os=linux
部署ingress-inginx controller
进入之前解压的folder (value.yaml)的上一层
执行
helm install nginx
gateman@MoreFine-S500:~/projects/coding/k8s-s/ingress/ingress-nginx/helm/ingress-nginx$ helm install ingress-nginx -n ingress-nginx .
NAME: ingress-nginx
LAST DEPLOYED: Tue Jun 25 02:16:57 2024
NAMESPACE: ingress-nginx
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
The ingress-nginx controller has been installed.
检查pods
gateman@MoreFine-S500:~/tmp/ingress-nginx$ kubectl get po -o wide -n ingress-nginx
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
ingress-nginx-controller-72dmb 1/1 Running 3 (2d18h ago) 6d20h 192.168.0.6 k8s-node0 <none> <none>
可见这个ingress-controller 已经运行在了 k8s-node0
创建1个ingress 资源
创建yaml
yaml file:
apiVersion: networking.k8s.io/v1
kind: Ingress # it is a resource name of k8s
metadata