背景:

接触kubernetes也好多年了，开始就各种听说Operator的，但是从来没有深入了解动手写过Operator。开始体验一下简单的Operator

Operator初体验

什么是Operator?

参照：红帽官方文档什么是 Kubernetes Operator？
**coreos2016年引入,**是一种封装、部署和管理 Kubernetes 应用的方法

1. crd webhook controller

开发工具:

what is crd

**CRD **全称是 Custom Resource Definition, CRD是一种无需编码就可以扩展原生kubenetes API接口的方式。适合扩展kubernetes的自定义接口和功能。如果想更为灵活的添加逻辑就需要API Aggregation方式.

开始准备

常用的开发工具有一下几种：

我的开发工具 包括goland kubebuilder kustomize，kubernetes1.23.6，工作环境rocky linux 8.5 go 1.17
注意：一定看一下go 版本 与开发工具对应版本，以及与kubernetes的版本

kubebuilder kustomize install

https://github.com/kubernetes-sigs/kubebuilder/releases

[root@zhangpeng ~]# wget https://github.com/kubernetes-sigs/kubebuilder/releases/download/v3.5.0/kubebuilder_linux_amd64
1

[root@zhangpeng ~]# mv kubebuilder_linux_amd64 /usr/bin/kubebuilder
[root@zhangpeng ~]# chmod +x /usr/bin/kubebuilder
[root@zhangpeng ~]# kubebuilder version
Version: main.version{KubeBuilderVersion:"3.5.0", KubernetesVendor:"1.24.1", GitCommit:"26d12ab1134964dbbc3f68877ebe9cf6314e926a", BuildDate:"2022-06-24T12:17:52Z", GoOs:"linux", GoArch:"amd64"}
1
2
3
4

root@zhangpeng ~]# wget https://github.com/kubernetes-sigs/kustomize/releases/download/kustomize%2Fv4.5.5/kustomize_v4.5.5_linux_amd64.tar.gz

1
2

[root@zhangpeng ~]# tar zxvf kustomize_v4.5.5_linux_amd64.tar.gz 
kustomize
[root@zhangpeng ~]# chmod +x kustomize
[root@zhangpeng ~]# mv kustomize /usr/bin/kustomize
[root@zhangpeng ~]# kustomize version
{Version:kustomize/v4.5.5 GitCommit:daa3e5e2c2d3a4b8c94021a7384bfb06734bcd26 BuildDate:2022-05-20T20:25:40Z GoOs:linux GoArch:amd64}
1
2
3
4
5
6

创建并初始化项目

goland创建名为kube-oprator1的项目：

终端执行一下命令：

[zhangpeng@zhangpeng kube-oprator1]$ kubebuilder init --plugins go/v3 --domain zhangpeng.com --owner "zhang peng"

1
2

貌似提示我go版本过低？（go版本 1.17.6我的是）

升级一下go版本

注意：非必须，后面是降低了kubebuilder的版本。go版本就保持1.17版本了，

浏览器打开https://golang.google.cn/dl/ go下载页面，选择1.17最新版本下载并替换本地GO版本！

[root@zhangpeng ~]# wget https://golang.google.cn/dl/go1.17.11.linux-amd64.tar.gz
[root@zhangpeng ~]# tar zxvf go1.17.11.linux-amd64.tar.gz
[root@zhangpeng ~]# which go
/usr/go/bin/go
[root@zhangpeng ~]# cd go/
[root@zhangpeng ~]# cp -Ra * /usr/go/
[root@zhangpeng go]# go version go1.17.11 linux/amd64
1
2
3
4
5
6
7

。。。。貌似还是报错，仔细看了一眼https://github.com/kubernetes-sigs/kubebuilder/releases我还是换个kubebuilder版本吧…

kubebuilder 版本3.4.1

[root@zhangpeng ~]# wget https://github.com/kubernetes-sigs/kubebuilder/releases/download/v3.4.1/kubebuilder_linux_amd64
[root@zhangpeng ~]# mv kubebuilder_linux_amd64 /usr/bin/kubebuilder
mv：是否覆盖'/usr/bin/kubebuilder'？ y
[root@zhangpeng ~]# chmod +x /usr/bin/kubebuilder 
[root@zhangpeng ~]# kubebuilder version
Version: main.version{KubeBuilderVersion:"3.4.1", KubernetesVendor:"1.23.5", GitCommit:"d59d7882ce95ce5de10238e135ddff31d8ede026", BuildDate:"2022-05-06T13:58:56Z", GoOs:"linux", GoArch:"amd64"}
1
2
3
4
5
6

kubebuilder init --plugins go/v3 --domain zhangpeng.com --owner "zhang peng"
1

生成目录结构如下：

重点关注一下config/default/ kustomization.yaml文件：

现在能看懂的配置，命名空间！我这里就不修改了采用默认的配置！

 [zhangpeng@zhangpeng kube-oprator1]$ kubebuilder create api --group myapp1 --version v1 --kind Redis

1
2

目录结构如下

注意:关于 domain group version kind对应 :

apiVersion:myapp1.zhangpeng.com/v1
kind: Redis
1
2

简单创建一个crd

api/v1/redis_type.go


随手演示删除Foo字段，添加一个Port字段，设置port字段为int类型！

以test目录下yaml文件定制crd

test/redis.yaml

apiVersion: myapp1.zhangpeng.com/v1
kind: Redis
metadata:
  name: myapp
spec:
  port: 1011
1
2
3
4
5
6

make install创建crd

[zhangpeng@zhangpeng kube-oprator1]$ kubectl get crd
No resources found
[zhangpeng@zhangpeng kube-oprator1]$ make install 
GOBIN=/home/zhangpeng/GolandProjects/kube-oprator1/bin go install sigs.k8s.io/controller-tools/cmd/controller-gen@v0.8.0
/home/zhangpeng/GolandProjects/kube-oprator1/bin/controller-gen rbac:roleName=manager-role crd webhook paths="./..." output:crd:artifacts:config=config/crd/bases
/home/zhangpeng/GolandProjects/kube-oprator1/bin/kustomize build config/crd | kubectl apply -f -
customresourcedefinition.apiextensions.k8s.io/redis.myapp1.zhangpeng.com created
[zhangpeng@zhangpeng kube-oprator1]$ kubectl get crd
NAME                         CREATED AT
redis.myapp1.zhangpeng.com   2022-06-28T06:44:52Z

1
2
3
4
5
6
7
8
9
10
11

关于reconcile

controllers/redis_controller.go
关于reconcile就先不求甚解了

func (r *RedisReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.Result, error) {
	_ = log.FromContext(ctx)

	// TODO(user): your logic here
	redis := &myapp1v1.Redis{}
	if err := r.Get(ctx, req.NamespacedName, redis); err != nil {
		fmt.Println(err)
	} else {
		fmt.Println("object", redis)
	}

	return ctrl.Result{}, nil
}
1
2
3
4
5
6
7
8
9
10
11
12
13

本地调试 make run

终端一运行

[zhangpeng@zhangpeng kube-oprator1]$ maker run  
1

终端2运行

 
 [zhangpeng@zhangpeng kube-oprator1]$ kubectl apply -f test/redis.yaml
1
2

观察终端1 得到如下输出：

初步发布到kubernetes集群

注：我的环境安装了podman，关于podman自行百度，镜像仓库使用了腾讯云镜像仓库个人版

关于Podman

先修改docker构建命令为podman!

podman login将密码记住…基本跟docker的使用方法差不多

[zhangpeng@zhangpeng kube-oprator1]$ podman login --username=xxxxx ccr.ccs.tencentyun.com

1
2

dockerhub加速

的特别强调一下dockerhub加速

[zhangpeng@zhangpeng kube-oprator1]$  vim /etc/containers/registries.conf
1

文件末尾添加了加速地址！

short-name-mode = "permissive"
[[registry]]
prefix="docker.io"
location="pvurwzu6.mirror.aliyuncs.com"
1
2
3
4

重启podman服务

[zhangpeng@zhangpeng kube-oprator1]$  systemctl restart podman
1

构建发布镜像

Dockerfile文件中添加GOPROXY

ENV GOPROXY=https://goproxy.io
1

[zhangpeng@zhangpeng kube-oprator1]$ make docker-build docker-push IMG=ccr.ccs.tencentyun.com/layatools/zpredis:v1
1

顺便发现腾讯云个人仓库一个不显示OCI-Image大小的bug…

注：过程很曲折。中间有镜像下不动的科学上网了，自行脑部。如“gcr.io/distroless/static:nonroot镜像我的操作环境为rocky linux 8.5下载不动的时候我直接科学上网了…

发布方式：

[zhangpeng@zhangpeng kube-oprator1]$ make deploy IMG=ccr.ccs.tencentyun.com/layatools/zpredis:v1
1

make又失败了最终根据Makefile中deploy手动执行如下命令：

[zhangpeng@zhangpeng kube-oprator1]$ cd config/manager &&  kustomize edit set image controller=ccr.ccs.tencentyun.com/layatools/zpredis:v1
[zhangpeng@zhangpeng kube-oprator1]$ kustomize build config/default | kubectl apply -f -

1
2
3

注意：两条命令都是在kube-oprator1项目根目录下执行的！


默认命名空间没有修改查看kube-oprator1-system namespace命名空间下pod状态！

[zhangpeng@zhangpeng kube-oprator1]$ kubectl get ns
NAME                   STATUS   AGE
default                Active   61d
kube-node-lease        Active   61d
kube-oprator1-system   Active   25h
kube-public            Active   61d
kube-system            Active   61d
zhangpeng1             Active   8d
[zhangpeng@zhangpeng kube-oprator1]$ kubectl get pods -n kube-oprator1-system

1
2
3
4
5
6
7
8
9
10

理论上pod是没有部署成功的，原因如下：gcr.io/kubebuilder/kube-rbac-proxy:v0.11.0无法下载…我是用的苯方法，rocky开发机科学上网下载镜像上传到腾讯与镜像仓库，然后pull镜像到kuberntes机器work节点。当然了work节点我只有一个测试环境还好～

等待pod running

CRD自定义资源简单验证

以test/redis.yaml为例

apiVersion: myapp1.zhangpeng.com/v1
kind: Redis
metadata:
  name: myapp
spec:
  port: 1011
1
2
3
4
5
6

参照https://book.kubebuilder.io/reference/markers/crd-validation.html 就设置一下port的范围！

	// +kubebuilder:validation:Minimum:=1024
	// +kubebuilder:validation:Maximum:=10240
1
2


make install 依然失败！还是手动命令了
注：这个地方我做错了好几次，原因是我以为make install =kustomize build config/crd | kubectl apply -f -,仔细看了一眼Makefile：
包含manifests的步骤，尝试了一下果然如此！

[zhangpeng@zhangpeng kube-oprator1]$ ./bin/controller-gen rbac:roleName=manager-role crd webhook paths="./..." output:crd:artifacts:config=config/crd/bases

[zhangpeng@zhangpeng kube-oprator1]$ kustomize build config/crd | kubectl apply -f -
1
2
3


[zhangpeng@zhangpeng kube-oprator1]$ kubectl get crd redis.myapp1.zhangpeng.com -o yaml

1
2


[zhangpeng@zhangpeng kube-oprator1]$ kubectl delete -f test/redis.yaml
[zhangpeng@zhangpeng kube-oprator1]$ kubectl apply -f test/redis.yaml
1
2

得到如下输出：端口小于1024无法创建成功
修改 test/redis.yaml port: 1024

apiVersion: myapp1.zhangpeng.com/v1
kind: Redis
metadata:
  name: myapp
spec:
  port: 1024
1
2
3
4
5
6


[zhangpeng@zhangpeng cert]$ kubectl get pods -A|grep cert
cert-manager           cert-manager-677874db78-zcm6l                       1/1     Running            0                  14m
cert-manager           cert-manager-cainjector-6c5bf7b759-mf4gf            1/1     Running            0                  14m
cert-manager           cert-manager-webhook-5685fdbc4b-ncrxl               1/1     Running            0                  14m

1
2
3
4
5

webhook简单测试

简单准入控制器webhook create

[zhangpeng@zhangpeng kube-oprator1]$ kubebuilder create webhook --group myapp1 --version v1 --kind Redis --defaulting --programmatic-validation

1
2

kube-oprator1 api/v1目录下增加了webhook的相关文件,做了一个简单的验证** name=zhangpeng**

func (r *Redis) ValidateCreate() error {
	redislog.Info("validate create", "name", r.Name)
	if r.Name == "zhangpeng" {
		return errors.New("error name")
	}

	// TODO(user): fill in your validation logic upon object creation.
	return nil
}
1
2
3
4
5
6
7
8
9

证书管理cert-manager:

访问https://github.com/cert-manager/cert-manager/releases下载页面，1.19.0是alpha版本 我用了1.18.2的版本！

[zhangpeng@zhangpeng cert]$ pwd
/home/zhangpeng/cert
[zhangpeng@zhangpeng cert]$ wget https://github.com/cert-manager/cert-manager/releases/download/v1.8.2/cert-manager.yaml
[zhangpeng@zhangpeng cert]$ kubectl apply -f cert-manager.yaml
[zhangpeng@zhangpeng cert]$ kubectl get pods -A|grep cert
1
2
3
4
5

注意：镜像依然要科学上网下载

修改文件

config/default/kustomization.yaml下箭头标注部分解开注释
**config/manager/manager.yaml **

删除crd

make uninstall可以不过我的make总是失败…直接删除了！

[zhangpeng@zhangpeng kube-oprator1]$kubectl delete crd redis.myapp1.zhangpeng.com

1
2


打包镜像发布

打包发布镜像，其实最好应该修改一个镜像标签tag，这里就演示 就先这样了！make install 也不能用不知道那里有问题了 直接复制Makefile中的命令了！构建镜像并发布镜像！

[zhangpeng@zhangpeng kube-oprator1]$ ./bin/controller-gen rbac:roleName=manager-role crd webhook paths="./..." output:crd:artifacts:config=config/crd/bases
[zhangpeng@zhangpeng kube-oprator1]$ kustomize build config/crd  | kubectl apply -f -
customresourcedefinition.apiextensions.k8s.io/redis.myapp1.zhangpeng.com configured

[zhangpeng@zhangpeng kube-oprator1]$ make docker-build docker-push IMG=ccr.ccs.tencentyun.com/layatools/zpredis:v1
[zhangpeng@zhangpeng kube-oprator1]$ make deploy IMG=ccr.ccs.tencentyun.com/layatools/zpredis:v1
1
2
3
4
5
6

[zhangpeng@zhangpeng kube-oprator1]$ cd config/manager && kustomize edit set image controller=ccr.ccs.tencentyun.com/layatools/zpredis:v1
[zhangpeng@zhangpeng kube-oprator1]$ kustomize build config/default | kubectl apply -f -
1
2

恩修改成zhangpeng1

创建成功了？看一下make run，but make run无法运行了？

本地调试模式注释掉main.go SetupWebhookWithManager

总结：

1.注意开发工具之间版本的匹配
2.make 失败时候看一下Makefile中相关命令可以手动运行一下
3.资源的清理，本地调试模式
4.接下来准备设计一个简单的oprator?