kubernetes 估计会成为 linux 一样的存在,client-go 是它的 go sdk,client-go/examples/ 给出了一些用例,但是数量比较少。
Resource 的定义不在client-go中,而是在一个名为 api 的项目中,这个项目中的内容同步自 kubernetes 项目的目录 staging/src/k8s.io/api,可以用下面的方式获取:
go get k8s.io/api
api的目录结构如下:
▾ api/
▸ admission/
▸ admissionregistration/
▸ apps/
▸ authentication/
▸ authorization/
▸ autoscaling/
▸ batch/
▸ certificates/
▸ core/
▸ extensions/
▸ Godeps/
▸ imagepolicy/
▸ networking/
▸ policy/
▸ rbac/
▸ scheduling/
▸ settings/
▸ storage/
▸ vendor/
LICENSE
OWNERS
README.md
client-go是kubernetes官方的项目,go语言实现,使用示例源码位于: go-code-example/k8sclient。
获取 client-go:
go get k8s.io/client-go
go get k8s.io/apimachinery
go get k8s.io/api
go get k8s.io/kube-openapi
Clientset 包含 kubernetes 的所有资源的操作句柄,通过”k8s.io/client-go/kubernetes”中NewForConfig()
创建。
创建 Clientset 需要提供一个 rest.Config,Config 可以用后面提到的 tools/clientcmd 生成,也可以自己填写:
config := rest.Config{
Host: "https://10.39.0.105:6443",
APIPath: "/",
Prefix: "",
BearerToken: "af8cbdf725efadf8c4",
TLSClientConfig: rest.TLSClientConfig{Insecure: true},
}
创建方法很简单:
clientset, err := kubernetes.NewForConfig(&config)
Clientset,意如其名,是client 的集合,在client-go/kubernetes/clientset.go中定义。
type Clientset struct {
*discovery.DiscoveryClient
admissionregistrationV1alpha1 *admissionregistrationv1alpha1.AdmissionregistrationV1alpha1Client
appsV1beta1 *appsv1beta1.AppsV1beta1Client
appsV1beta2 *appsv1beta2.AppsV1beta2Client
appsV1 *appsv1.AppsV1Client
authenticationV1 *authenticationv1.AuthenticationV1Client
authenticationV1beta1 *authenticationv1beta1.AuthenticationV1beta1Client
authorizationV1 *authorizationv1.AuthorizationV1Client
authorizationV1beta1 *authorizationv1beta1.AuthorizationV1beta1Client
autoscalingV1 *autoscalingv1.AutoscalingV1Client
autoscalingV2beta1 *autoscalingv2beta1.AutoscalingV2beta1Client
batchV1 *batchv1.BatchV1Client
batchV1beta1 *batchv1beta1.BatchV1beta1Client
batchV2alpha1 *batchv2alpha1.BatchV2alpha1Client
certificatesV1beta1 *certificatesv1beta1.CertificatesV1beta1Client
coreV1 *corev1.CoreV1Client
extensionsV1beta1 *extensionsv1beta1.ExtensionsV1beta1Client
networkingV1 *networkingv1.NetworkingV1Client
policyV1beta1 *policyv1beta1.PolicyV1beta1Client
rbacV1 *rbacv1.RbacV1Client
rbacV1beta1 *rbacv1beta1.RbacV1beta1Client
rbacV1alpha1 *rbacv1alpha1.RbacV1alpha1Client
schedulingV1alpha1 *schedulingv1alpha1.SchedulingV1alpha1Client
settingsV1alpha1 *settingsv1alpha1.SettingsV1alpha1Client
storageV1beta1 *storagev1beta1.StorageV1beta1Client
storageV1 *storagev1.StorageV1Client
}
用 Clientset 的方法获得不同资源的操作句柄,Clientset有以下的方法可供使用:
+Admissionregistration() : admissionregistrationv1alpha1.AdmissionregistrationV1alpha1Interface
+AdmissionregistrationV1alpha1() : admissionregistrationv1alpha1.AdmissionregistrationV1alpha1Interface
+Apps() : appsv1.AppsV1Interface
+AppsV1() : appsv1.AppsV1Interface
+AppsV1beta1() : appsv1beta1.AppsV1beta1Interface
+AppsV1beta2() : appsv1beta2.AppsV1beta2Interface
+Authentication() : authenticationv1.AuthenticationV1Interface
+AuthenticationV1() : authenticationv1.AuthenticationV1Interface
+AuthenticationV1beta1() : authenticationv1beta1.AuthenticationV1beta1Interface
+Authorization() : authorizationv1.AuthorizationV1Interface
+AuthorizationV1() : authorizationv1.AuthorizationV1Interface
+AuthorizationV1beta1() : authorizationv1beta1.AuthorizationV1beta1Interface
+Autoscaling() : autoscalingv1.AutoscalingV1Interface
+AutoscalingV1() : autoscalingv1.AutoscalingV1Interface
+AutoscalingV2beta1() : autoscalingv2beta1.AutoscalingV2beta1Interface
+Batch() : batchv1.BatchV1Interface
+BatchV1() : batchv1.BatchV1Interface
+BatchV1beta1() : batchv1beta1.BatchV1beta1Interface
+BatchV2alpha1() : batchv2alpha1.BatchV2alpha1Interface
+Certificates() : certificatesv1beta1.CertificatesV1beta1Interface
+CertificatesV1beta1() : certificatesv1beta1.CertificatesV1beta1Interface
+Core() : corev1.CoreV1Interface
+CoreV1() : corev1.CoreV1Interface
+Discovery() : discovery.DiscoveryInterface
+Extensions() : extensionsv1beta1.ExtensionsV1beta1Interface
+ExtensionsV1beta1() : extensionsv1beta1.ExtensionsV1beta1Interface
+Networking() : networkingv1.NetworkingV1Interface
+NetworkingV1() : networkingv1.NetworkingV1Interface
+Policy() : policyv1beta1.PolicyV1beta1Interface
+PolicyV1beta1() : policyv1beta1.PolicyV1beta1Interface
+Rbac() : rbacv1.RbacV1Interface
+RbacV1() : rbacv1.RbacV1Interface
+RbacV1alpha1() : rbacv1alpha1.RbacV1alpha1Interface
+RbacV1beta1() : rbacv1beta1.RbacV1beta1Interface
+Scheduling() : schedulingv1alpha1.SchedulingV1alpha1Interface
+SchedulingV1alpha1() : schedulingv1alpha1.SchedulingV1alpha1Interface
+Settings() : settingsv1alpha1.SettingsV1alpha1Interface
+SettingsV1alpha1() : settingsv1alpha1.SettingsV1alpha1Interface
+Storage() : storagev1.StorageV1Interface
+StorageV1() : storagev1.StorageV1Interface
+StorageV1beta1() : storagev1beta1.StorageV1beta1Interface
其中Core()和CoreV1()获取到的corev1.CoreV1Interface,用来操作kubernetes的最基础的Resrouce。
type CoreV1Interface interface {
RESTClient() rest.Interface
ComponentStatusesGetter
ConfigMapsGetter
EndpointsGetter
EventsGetter
LimitRangesGetter
NamespacesGetter
NodesGetter
PersistentVolumesGetter
PersistentVolumeClaimsGetter
PodsGetter
PodTemplatesGetter
ReplicationControllersGetter
ResourceQuotasGetter
SecretsGetter
ServicesGetter
ServiceAccountsGetter
}
例如查找指定 namespace 中的所有 Pod:
pods, err := clientset.CoreV1().Pods("lijiaocn").List(v1.ListOptions{})
client-go/rest 实现了 RESTClient,RESTClient 的创建方法有两种。
第一种方式是直接通过 config 创建,这种方式需要在 config 中填入 GroupVersion 信息:
// create restClient from config
func CreateCoreRestClient(config *rest.Config) (*rest.RESTClient, error) {
config.ContentConfig.GroupVersion = &core_v1.SchemeGroupVersion
config.ContentConfig.NegotiatedSerializer = serializer.DirectCodecFactory{CodecFactory: scheme.Codecs}
config.APIPath = "/api"
restClient, err := rest.RESTClientFor(config)
if err != nil {
return nil, err
}
if restClient == nil {
return nil, errors.New("restclient1 is nil")
}
return restClient, nil
}
第二种方式是从 Clientset 中获取,这种方式不需要手动填入 GroupVersion,调用 Clientset 对应的接口即可:
// get restClient from clientset
func GetCoreRestClient(config *rest.Config) (rest.Interface, error) {
clientset, err := kubernetes.NewForConfig(config)
if err != nil {
return nil, err
}
if clientset == nil {
return nil, errors.New("clientset is nil")
}
restClient := clientset.CoreV1().RESTClient()
if restClient == nil {
return nil, errors.New("restclient is nil")
}
return restClient, nil
}
RESTClient 是链式调用,使用方法如下,Do() 方法最后调用:
// /<namespace>/<resource>/<name>
// GET https://10.10.173.203/api/v1/namespaces/default/pods?limit=500
// GET https://10.10.173.203/api/v1/namespaces/kube-system/pods?limit=500
// GET https://10.10.173.203/api/v1/namespaces/kube-system/pods/kube-dns-5b54cf547c-jl4r9
result := restClient.Get().
Namespace("kube-system").
Resource("pods").
Name("kube-dns-5b54cf547c-jl4r9").
Do()
bytes, err := result.Raw()
if err != nil {
fmt.Printf("%s: %s\n", err.Error(), bytes)
} else {
fmt.Printf("%s\n", bytes)
}
client-go 提供一套 tools,提供了配置文件加载、实现本地缓存等方法。
tools/clientcmd 中提供了一些辅助工具,例如加载 kubeconfig 文件生成创建 client 是必须的 Config:
home, err := os.UserHomeDir()
if err != nil {
glog.Fatal(err.Error())
}
file, err := filepath.Abs(home + "/.kube/config")
if err != nil {
glog.Fatal(err.Error())
}
config, err := clientcmd.BuildConfigFromFlags("", file)
if err != nil {
glog.Fatal(err.Error())
return
}
clientset, err := kubernetes.NewForConfig(config)
if err != nil {
glog.Fatal(err.Error())
return
}
tools/cache 中提供了本地缓存的功能,特别是提供了 informer。
创建 informer 需要提供的 listwatcher 和 handler 都用 cache 中方法创建或在 cache 中定义:
labels := make(map[string]string)
selector := fields.SelectorFromSet(labels)
listwatch := cache.NewListWatchFromClient(restClient, "endpoints", "", selector)
handler 定义:
handler := cache.ResourceEventHandlerFuncs{
AddFunc: func(obj interface{}) {
fmt.Printf("Add endpoint:\n")
if ep, ok := obj.(*core_v1.Endpoints); !ok {
fmt.Printf("not endpoints\n")
} else {
printEndpoint(ep)
}
},
UpdateFunc: func(oldObj, newObj interface{}) {
fmt.Printf("Update endpoint:\n")
if epOld, ok := oldObj.(*core_v1.Endpoints); !ok {
fmt.Printf("not endpoints\n")
} else {
printEndpoint(epOld)
}
if epNew, ok := newObj.(*core_v1.Endpoints); !ok {
fmt.Printf("not endpoints\n")
} else {
printEndpoint(epNew)
}
},
DeleteFunc: func(obj interface{}) {
fmt.Printf("Delete endpoint:\n")
if ep, ok := obj.(*core_v1.Endpoints); !ok {
fmt.Printf("not endpoint")
} else {
printEndpoint(ep)
}
},
}
cache.NewInformer() 返回一个 store,一个controller,前者用于查询,后者用于控制同步。
stop := make(chan struct{})
store, controller := cache.NewInformer(listwatch, &core_v1.Endpoints{}, 0*time.Second, handler)
controller.Run(stop)