kubernetes CRD開發指南

擴展kubernetes兩個最常用最需要掌握的東西：自定義資源CRD 和 adminsion webhook, 本文教你如何十分鐘掌握CRD開發.

kubernetes允許用戶自定義自己的資源對象，就如同deployment statefulset一樣，這個應用非常廣泛，比如prometheus opterator就自定義Prometheus對象，再加上一個自定義的controller監聽到kubectl create Prometheus時就去創建Pod組成一個pormetheus集羣。rook等等同理。

我需要用kubernetes調度虛擬機，所以這裏自定義一個 VirtualMachine 類型
<!--more-->

kubebuilder

kubebuilder能幫我們節省大量工作，讓開發CRD和adminsion webhook變得異常簡單。

安裝

通過源碼安裝：

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

或者下載二進制：

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

還需要裝下kustomize 這可是個渲染yaml的神器，讓helm顫抖。

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

使用

注意你得先有個kubernetes集羣，一步安裝走你

創建CRD

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

安裝CRD並啓動controller

make install # 安裝CRD
make run # 啓動controller

然後我們就可以看到創建的CRD了

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

來創建一個虛擬機：

# 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

這裏僅僅是把yaml存到etcd裏了，我們controller監聽到創建事件時啥事也沒幹。

把controller部署到集羣中

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

我是連的遠端的kubenetes, make docker-build時test過不去，沒有etcd的bin文件，所以先把test關了。

修改Makefile:

# docker-build: test
docker-build: 

Dockerfile裏的gcr.io/distroless/static:latest 這個鏡像你也可能拉不下來，隨意改改就行，我改成了golang:1.12.7

也有可能構建時有些代碼拉不下來，啓用一下go mod vendor 把依賴打包進去

go mod vendor
如果你本地有些代碼拉不下來，可以用proxy:

export GOPROXY=https://goproxy.io

再改下Dockerfile, 註釋掉download：

修改後：

# 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 時報錯： 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 這個命令就渲染出了controller的yaml文件，可以體驗下

看 你的controller已經跑起來了：

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   <none>        8443/TCP   4m

開發

增加對象數據參數

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

這裏參數裏有foo:bar， 那我們來加個虛擬CPU，內存信息：

直接api/v1/virtulmachine_types.go即可

// VirtulMachineSpec defines the desired state of VirtulMachine
// 在這裏加信息
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"`   // 這是我增加的
    Memory string `json:"memory"`
}

// VirtulMachineStatus defines the observed state of VirtulMachine
// 在這裏加狀態信息，比如虛擬機是啓動狀態，停止狀態啥的
type VirtulMachineStatus struct {
    // INSERT ADDITIONAL STATUS FIELD - define observed state of cluster
    // Important: Run "make" to regenerate code after modifying this file
}

然後make一下：

make && make install && make run

這時再去渲染一下controller的yaml就會發現CRD中已經帶上CPU和內存信息了：

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，就不再贅述了，比如我把status里加一個Create, 表示controller要去創建虛擬機了(主要一些控制層面的邏輯)，創建完了把狀態改成Running

Reconcile 唯一需要實現的接口

controller把輪訓與事件監聽都封裝在這一個接口裏了.你不需要關心怎麼事件監聽的.

獲取虛擬機信息

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 { # 獲取VM信息
        log.Error(err, "unable to fetch vm")
    } else {
        fmt.Println(vm.Spec.CPU, vm.Spec.Memory) # 打印CPU內存信息
    }

    return ctrl.Result{}, nil
}

make && make install && make run這個時候去創建一個虛擬機kubectl apply -f config/samples,日誌裏就會輸出CPU內存了. List接口同理，我就不贅述了

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

更新狀態

在status結構體中加入狀態字段：

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

controller裏去更新狀態：

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

如果出現:the server could not find the requested resource 這個錯誤，那麼在CRD結構體上需要加個註釋 // +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"`
}

這樣就好了

編譯啓動後再去apply發現狀態已經變成running：

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

刪除

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

10s之後我們將GET不到

刪除回收器 Finalizers

如果不使用Finalizers，kubectl delete 時直接就刪了etcd數據，controller再想去拿CRD時已經拿不到了：

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

所以在創建時我們需要給CRD加上Finalizer:

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

然後刪除時就只會給CRD打上一個刪除時間戳，供我們做後續處理, 處理完了我們刪除掉Finalizers：

如果 DeleteionTimestamp不存在
    如果沒有Finalizers
        加上Finalizers,並更新CRD
要不然，說明是要被刪除的
    如果存在Finalizers，刪除Finalizers,並更新CRD

看個完整的代碼示例：

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
            }
        }
    }

失敗重試

假設我們A依賴B而B又後創建，那麼在處理A CRD時直接返回失敗即可，這樣很快會重試

webhook

kuberentes有三種webhook,admission webhook, authorization webhook and CRD conversion webhook.

這裏比如我們要給CRD設置一些默認值，又或者是用戶創建時少填了一些參數，那麼我們得禁止創建等等這些事。

使用webhook也非常的簡單，只需給定義的結構體實現 Defaulter 和 Validator接口即可.

其它接口

Reconcile結構體聚合了Client接口，所以client的所有方法都是可以直接調用，大部分是對CRD object的相關操作

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
}

掃碼關注sealyun

探討可加QQ羣：98488045