kubernetes CRD开发指南

扩展kubernetes两个最常用最需要掌握的东西：自定义资源CRD 和 adminsion webhook, 本文教你如何十分钟掌握CRD开发.

kubernetes允许用户自定义自己的资源对象，就如同deployment statefulset一样，这个应用非常广泛，比如prometheus opterator就自定义Prometheus对象，再加上一个自定义的controller监听到kubectl create Prometheus时就去创建Pod组成一个pormetheus集群。rook等等同理。

我需要用kubernetes调度虚拟机，所以这里自定义一个 VirtualMachine 类型

kubebuilder

kubebuilder能帮我们节省大量工作，让开发CRD和adminsion webhook变得异常简单。

安装

通过源码安装：

git clone https://github.com/kubernetes-sigs/kubebuilder
cd kubebuilder
make build
cp bin/kubebuilder $GOPATH/bin

或者下载二进制：

os=$(go env GOOS)
arch=$(go env GOARCH)

# download kubebuilder and extract it to tmp
curl -sL https://go.kubebuilder.io/dl/2.0.0-beta.0/${os}/${arch} | tar -xz -C /tmp/

# move to a long-term location and put it on your path
# (you'll need to set the KUBEBUILDER_ASSETS env var if you put it somewhere else)
sudo mv /tmp/kubebuilder_2.0.0-beta.0_${os}_${arch} /usr/local/kubebuilder
export PATH=$PATH:/usr/local/kubebuilder/bin

还需要装下kustomize 这可是个渲染yaml的神器，让helm颤抖。

go install sigs.k8s.io/kustomize/v3/cmd/kustomize

使用

注意你得先有个kubernetes集群，一步安装走你

创建CRD

kubebuilder init --domain sealyun.com --license apache2 --owner "fanux"
kubebuilder create api --group infra --version v1 --kind VirtulMachine

安装CRD并启动controller

make install # 安装CRD
make run # 启动controller

然后我们就可以看到创建的CRD了

# kubectl get crd
NAME                                           AGE
virtulmachines.infra.sealyun.com                  52m

来创建一个虚拟机：

# kubectl apply -f config/samples/
# kubectl get virtulmachines.infra.sealyun.com 
NAME                   AGE
virtulmachine-sample   49m

看一眼yaml文件：

# cat config/samples/infra_v1_virtulmachine.yaml 
apiVersion: infra.sealyun.com/v1
kind: VirtulMachine
metadata:
  name: virtulmachine-sample
spec:
  # Add fields here
  foo: bar

这里仅仅是把yaml存到etcd里了，我们controller监听到创建事件时啥事也没干。

把controller部署到集群中

make docker-build docker-push IMG=fanux/infra-controller
make deploy

我是连的远端的kubenetes, make docker-build时test过不去，没有etcd的bin文件，所以先把test关了。

修改Makefile:

# docker-build: test
docker-build: 

Dockerfile里的gcr.io/distroless/static:latest 这个镜像你也可能拉不下来，随意改改就行，我改成了golang:1.12.7

也有可能构建时有些代码拉不下来，启用一下go mod vendor 把依赖打包进去

go mod vendor
如果你本地有些代码拉不下来，可以用proxy:

export GOPROXY=https://goproxy.io

再改下Dockerfile, 注释掉download：

修改后：

# Build the manager binary
FROM golang:1.12.7 as builder

WORKDIR /go/src/github.com/fanux/sealvm
# Copy the Go Modules manifests
COPY . . 

# Build
RUN CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -o manager main.go

# Use distroless as minimal base image to package the manager binary
# Refer to https://github.com/GoogleContainerTools/distroless for more details
# FROM gcr.io/distroless/static:latest
FROM golang:1.12.7
WORKDIR /
COPY --from=builder /go/src/github.com/fanux/sealvm/manager .
ENTRYPOINT ["/manager"]

make deploy 时报错： Error: json: cannot unmarshal string into Go struct field Kustomization.patches of type types.Patch

把 config/default/kustomization.yaml 中的 patches: 改成 patchesStrategicMerge: 即可

kustomize build config/default 这个命令就渲染出了controller的yaml文件，可以体验下

看 你的controller已经跑起来了：

kubectl get deploy -n sealvm-system
NAME                        DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
sealvm-controller-manager   1         1         1            0           3m
kubectl get svc -n sealvm-system
NAME                                        TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)    AGE
sealvm-controller-manager-metrics-service   ClusterIP   10.98.71.199           8443/TCP   4m

开发

增加对象数据参数

看下config/samples下面的yaml文件：

apiVersion: infra.sealyun.com/v1
kind: VirtulMachine
metadata:
  name: virtulmachine-sample
spec:
  # Add fields here
  foo: bar

这里参数里有foo:bar， 那我们来加个虚拟CPU，内存信息：

直接api/v1/virtulmachine_types.go即可

// VirtulMachineSpec defines the desired state of VirtulMachine
// 在这里加信息
type VirtulMachineSpec struct {
    // INSERT ADDITIONAL SPEC FIELDS - desired state of cluster
    // Important: Run "make" to regenerate code after modifying this file
    CPU    string `json:"cpu"`   // 这是我增加的
    Memory string `json:"memory"`
}

// VirtulMachineStatus defines the observed state of VirtulMachine
// 在这里加状态信息，比如虚拟机是启动状态，停止状态啥的
type VirtulMachineStatus struct {
    // INSERT ADDITIONAL STATUS FIELD - define observed state of cluster
    // Important: Run "make" to regenerate code after modifying this file
}

然后make一下：

make && make install && make run

这时再去渲染一下controller的yaml就会发现CRD中已经带上CPU和内存信息了：

kustomize build config/default

properties:
  cpu:
    description: 'INSERT ADDITIONAL SPEC FIELDS - desired state of cluster
      Important: Run "make" to regenerate code after modifying this file'
    type: string
  memory:
    type: string

修改一下yaml:

apiVersion: infra.sealyun.com/v1
kind: VirtulMachine
metadata:
  name: virtulmachine-sample
spec:
  cpu: "1"
  memory: "2G"

# kubectl apply -f config/samples 
virtulmachine.infra.sealyun.com "virtulmachine-sample" configured
# kubectl get virtulmachines.infra.sealyun.com virtulmachine-sample -o yaml 
apiVersion: infra.sealyun.com/v1
kind: VirtulMachine
metadata:
  annotations:
    kubectl.kubernetes.io/last-applied-configuration: |
      {"apiVersion":"infra.sealyun.com/v1","kind":"VirtulMachine","metadata":{"annotations":{},"name":"virtulmachine-sample","namespace":"default"},"spec":{"cpu":"1","memory":"2G"}}
  creationTimestamp: 2019-07-26T08:47:34Z
  generation: 2
  name: virtulmachine-sample
  namespace: default
  resourceVersion: "14811698"
  selfLink: /apis/infra.sealyun.com/v1/namespaces/default/virtulmachines/virtulmachine-sample
  uid: 030e2b9a-af82-11e9-b63e-5254bc16e436
spec:      # 新的CRD已生效
  cpu: "1"
  memory: 2G 

Status 同理，就不再赘述了，比如我把status里加一个Create, 表示controller要去创建虚拟机了(主要一些控制层面的逻辑)，创建完了把状态改成Running

Reconcile 唯一需要实现的接口

controller把轮训与事件监听都封装在这一个接口里了.你不需要关心怎么事件监听的.

获取虚拟机信息

func (r *VirtulMachineReconciler) Reconcile(req ctrl.Request) (ctrl.Result, error) {
    ctx = context.Background()
    _ = r.Log.WithValues("virtulmachine", req.NamespacedName)

    vm := &v1.VirtulMachine{}
    if err := r.Get(ctx, req.NamespacedName, vm); err != nil { # 获取VM信息
        log.Error(err, "unable to fetch vm")
    } else {
        fmt.Println(vm.Spec.CPU, vm.Spec.Memory) # 打印CPU内存信息
    }

    return ctrl.Result{}, nil
}

make && make install && make run这个时候去创建一个虚拟机kubectl apply -f config/samples,日志里就会输出CPU内存了. List接口同理，我就不赘述了

r.List(ctx, &vms, client.InNamespace(req.Namespace), client.MatchingField(vmkey, req.Name))

更新状态

在status结构体中加入状态字段：

type VirtulMachineStatus struct {
    Status string `json:"status"`
}

controller里去更新状态：

vm.Status.Status = "Running"
if err := r.Status().Update(ctx, vm); err != nil {
    log.Error(err, "unable to update vm status")
}

如果出现:the server could not find the requested resource 这个错误，那么在CRD结构体上需要加个注释 // +kubebuilder:subresource:status：

// +kubebuilder:subresource:status
// +kubebuilder:object:root=true

type VirtulMachine struct {
    metav1.TypeMeta   `json:",inline"`
    metav1.ObjectMeta `json:"metadata,omitempty"`

    Spec   VirtulMachineSpec   `json:"spec,omitempty"`
    Status VirtulMachineStatus `json:"status,omitempty"`
}

这样就好了

编译启动后再去apply发现状态已经变成running：

# kubectl get virtulmachines.infra.sealyun.com virtulmachine-sample -o yaml
...
status:
  status: Running 

删除

time.Sleep(time.Second * 10)
if err := r.Delete(ctx, vm); err != nil {
    log.Error(err, "unable to delete vm ", "vm", vm)
}

10s之后我们将GET不到

删除回收器 Finalizers

如果不使用Finalizers，kubectl delete 时直接就删了etcd数据，controller再想去拿CRD时已经拿不到了：

ERRO[0029] VirtulMachine.infra.sealyun.com "virtulmachine-sample" not foundunable to fetch vm  source="virtulmachine_controller.go:48"

所以在创建时我们需要给CRD加上Finalizer:

vm.ObjectMeta.Finalizers = append(vm.ObjectMeta.Finalizers, "virtulmachine.infra.sealyun.com")

然后删除时就只会给CRD打上一个删除时间戳，供我们做后续处理, 处理完了我们删除掉Finalizers：

如果 DeleteionTimestamp不存在
    如果没有Finalizers
        加上Finalizers,并更新CRD
要不然，说明是要被删除的
    如果存在Finalizers，删除Finalizers,并更新CRD

看个完整的代码示例：

if cronJob.ObjectMeta.DeletionTimestamp.IsZero() {
        if !containsString(cronJob.ObjectMeta.Finalizers, myFinalizerName) {
            cronJob.ObjectMeta.Finalizers = append(cronJob.ObjectMeta.Finalizers, myFinalizerName)
            if err := r.Update(context.Background(), cronJob); err != nil {
                return ctrl.Result{}, err
            }
        }
    } else {
        if containsString(cronJob.ObjectMeta.Finalizers, myFinalizerName) {
            if err := r.deleteExternalResources(cronJob); err != nil {
                return ctrl.Result{}, err
            }

            cronJob.ObjectMeta.Finalizers = removeString(cronJob.ObjectMeta.Finalizers, myFinalizerName)
            if err := r.Update(context.Background(), cronJob); err != nil {
                return ctrl.Result{}, err
            }
        }
    }

失败重试

假设我们A依赖B而B又后创建，那么在处理A CRD时直接返回失败即可，这样很快会重试

webhook

kuberentes有三种webhook,admission webhook, authorization webhook and CRD conversion webhook.

这里比如我们要给CRD设置一些默认值，又或者是用户创建时少填了一些参数，那么我们得禁止创建等等这些事。

使用webhook也非常的简单，只需给定义的结构体实现 Defaulter 和 Validator接口即可.

其它接口

Reconcile结构体聚合了Client接口，所以client的所有方法都是可以直接调用，大部分是对CRD object的相关操作

type Client interface {
    Reader
    Writer
    StatusClient
}

// Reader knows how to read and list Kubernetes objects.
type Reader interface {
    // Get retrieves an obj for the given object key from the Kubernetes Cluster.
    // obj must be a struct pointer so that obj can be updated with the response
    // returned by the Server.
    Get(ctx context.Context, key ObjectKey, obj runtime.Object) error

    // List retrieves list of objects for a given namespace and list options. On a
    // successful call, Items field in the list will be populated with the
    // result returned from the server.
    List(ctx context.Context, list runtime.Object, opts ...ListOptionFunc) error
}

// Writer knows how to create, delete, and update Kubernetes objects.
type Writer interface {
    // Create saves the object obj in the Kubernetes cluster.
    Create(ctx context.Context, obj runtime.Object, opts ...CreateOptionFunc) error

    // Delete deletes the given obj from Kubernetes cluster.
    Delete(ctx context.Context, obj runtime.Object, opts ...DeleteOptionFunc) error

    // Update updates the given obj in the Kubernetes cluster. obj must be a
    // struct pointer so that obj can be updated with the content returned by the Server.
    Update(ctx context.Context, obj runtime.Object, opts ...UpdateOptionFunc) error

    // Patch patches the given obj in the Kubernetes cluster. obj must be a
    // struct pointer so that obj can be updated with the content returned by the Server.
    Patch(ctx context.Context, obj runtime.Object, patch Patch, opts ...PatchOptionFunc) error
}

// StatusClient knows how to create a client which can update status subresource
// for kubernetes objects.
type StatusClient interface {
    Status() StatusWriter
}

扫码关注sealyun

探讨可加QQ群：98488045