kubectl源码分析之top node

关注

原创最新推荐文章于 2026-09-20 13:49:43 发布 1.6k 阅读 AI 写同款· GEO 优化 ›

kubectl滚动更新资源 rollout Manage the rollout of a resource. Valid resource types include: * deployments * daemonsets * statefulsets Examples: # Rollback to the previous deployment kubectl rollout undo deployment/abc # Check the rollout status of a da 阅读详情

欢迎关注我的公众号：

目前刚开始写一个月，一共写了18篇原创文章，文章目录如下：

istio多集群探秘，部署了50次多集群后我得出的结论

istio多集群链路追踪，附实操视频

istio防故障利器，你知道几个,istio新手不要读，太难！

istio业务权限控制，原来可以这么玩

istio实现非侵入压缩，微服务之间如何实现压缩

不懂envoyfilter也敢说精通istio系列-http-rbac-不要只会用AuthorizationPolicy配置权限

不懂envoyfilter也敢说精通istio系列-02-http-corsFilter-不要只会vs

不懂envoyfilter也敢说精通istio系列-03-http-csrf filter-再也不用再代码里写csrf逻辑了

不懂envoyfilter也敢说精通istio系列http-jwt_authn-不要只会RequestAuthorization

不懂envoyfilter也敢说精通istio系列-05-fault-filter-故障注入不止是vs

不懂envoyfilter也敢说精通istio系列-06-http-match-配置路由不只是vs

不懂envoyfilter也敢说精通istio系列-07-负载均衡配置不止是dr

不懂envoyfilter也敢说精通istio系列-08-连接池和断路器

不懂envoyfilter也敢说精通istio系列-09-http-route filter

不懂envoyfilter也敢说精通istio系列-network filter-redis proxy

不懂envoyfilter也敢说精通istio系列-network filter-HttpConnectionManager

不懂envoyfilter也敢说精通istio系列-ratelimit-istio ratelimit完全手册

加qq群，请联系：

————————————————

//创建top命令
func NewCmdTop(f cmdutil.Factory, streams genericclioptions.IOStreams) *cobra.Command {
	cmd := &cobra.Command{//创建cobra命令
		Use:   "top",
		Short: i18n.T("Display Resource (CPU/Memory/Storage) usage."),
		Long:  topLong,
		Run:   cmdutil.DefaultSubCommandRun(streams.ErrOut),
	}

	// create subcommands
	cmd.AddCommand(NewCmdTopNode(f, nil, streams))//添加node子命令
	cmd.AddCommand(NewCmdTopPod(f, nil, streams))//添加pod子命令

	return cmd
}

type TopNodeOptions struct {//top node结构体
	ResourceName    string
	Selector        string
	SortBy          string
	NoHeaders       bool
	NodeClient      corev1client.CoreV1Interface
	HeapsterOptions HeapsterTopOptions
	Client          *metricsutil.HeapsterMetricsClient
	Printer         *metricsutil.TopCmdPrinter
	DiscoveryClient discovery.DiscoveryInterface
	MetricsClient   metricsclientset.Interface

	genericclioptions.IOStreams
}

type HeapsterTopOptions struct {//Heapster结构体
	Namespace string
	Service   string
	Scheme    string
	Port      string
}

func (o *HeapsterTopOptions) Bind(flags *pflag.FlagSet) {//设置heapster选项
	if len(o.Namespace) == 0 {
		o.Namespace = metricsutil.DefaultHeapsterNamespace
	}
	if len(o.Service) == 0 {
		o.Service = metricsutil.DefaultHeapsterService
	}
	if len(o.Scheme) == 0 {
		o.Scheme = metricsutil.DefaultHeapsterScheme
	}
	if len(o.Port) == 0 {
		o.Port = metricsutil.DefaultHeapsterPort
	}

	flags.StringVar(&o.Namespace, "heapster-namespace", o.Namespace, "Namespace Heapster service is located in")
	flags.StringVar(&o.Service, "heapster-service", o.Service, "Name of Heapster service")
	flags.StringVar(&o.Scheme, "heapster-scheme", o.Scheme, "Scheme (http or https) to connect to Heapster as")
	flags.StringVar(&o.Port, "heapster-port", o.Port, "Port name in service to use")
}

//创建top node命令
func NewCmdTopNode(f cmdutil.Factory, o *TopNodeOptions, streams genericclioptions.IOStreams) *cobra.Command {
	if o == nil {
		o = &TopNodeOptions{//初始化结构体
			IOStreams: streams,
		}
	}

	cmd := &cobra.Command{//创建cobra命令
		Use:                   "node [NAME | -l label]",
		DisableFlagsInUseLine: true,
		Short:                 i18n.T("Display Resource (CPU/Memory/Storage) usage of nodes"),
		Long:                  topNodeLong,
		Example:               topNodeExample,
		Run: func(cmd *cobra.Command, args []string) {
			cmdutil.CheckErr(o.Complete(f, cmd, args))//准备
			cmdutil.CheckErr(o.Validate())//校验
			cmdutil.CheckErr(o.RunTopNode())//运行
		},
		Aliases: []string{"nodes", "no"},// 别名
	}
	cmd.Flags().StringVarP(&o.Selector, "selector", "l", o.Selector, "Selector (label query) to filter on, supports '=', '==', and '!='.(e.g. -l key1=value1,key2=value2)")//selector选项
	cmd.Flags().StringVar(&o.SortBy, "sort-by", o.Selector, "If non-empty, sort nodes list using specified field. The field can be either 'cpu' or 'memory'.")//sort-by选项
	cmd.Flags().BoolVar(&o.NoHeaders, "no-headers", o.NoHeaders, "If present, print output without headers")//no-headers选项

	o.HeapsterOptions.Bind(cmd.Flags())//heapster选项
	return cmd
}

//准备
func (o *TopNodeOptions) Complete(f cmdutil.Factory, cmd *cobra.Command, args []string) error {
	if len(args) == 1 {
		o.ResourceName = args[0]//设置resource
	} else if len(args) > 1 {
		return cmdutil.UsageErrorf(cmd, "%s", cmd.Use)
	}

	clientset, err := f.KubernetesClientSet()//获取clientset
	if err != nil {
		return err
	}

	o.DiscoveryClient = clientset.DiscoveryClient//设置discoveryclient

	config, err := f.ToRESTConfig()//获取rest config
	if err != nil {
		return err
	}
	o.MetricsClient, err = metricsclientset.NewForConfig(config)//设置MetricsClient
	if err != nil {
		return err
	}

	o.NodeClient = clientset.CoreV1()//设置NodeClient 
	o.Client = metricsutil.NewHeapsterMetricsClient(clientset.CoreV1(), o.HeapsterOptions.Namespace, o.HeapsterOptions.Scheme, o.HeapsterOptions.Service, o.HeapsterOptions.Port)//设置Heapster client

	o.Printer = metricsutil.NewTopCmdPrinter(o.Out)//设置printer
	return nil
}

//校验函数
func (o *TopNodeOptions) Validate() error {
	if len(o.SortBy) > 0 {//排序
		if o.SortBy != sortByCPU && o.SortBy != sortByMemory {//只能是cpu或momery
			return errors.New("--sort-by accepts only cpu or memory")
		}
	}
	if len(o.ResourceName) > 0 && len(o.Selector) > 0 {//资源名称和selector不能同时指定
		return errors.New("only one of NAME or --selector can be provided")
	}
	return nil
}

//运行
func (o TopNodeOptions) RunTopNode() error {
	var err error
	selector := labels.Everything()//设置selector
	if len(o.Selector) > 0 {//如果--selector有值
		selector, err = labels.Parse(o.Selector)//设置selector
		if err != nil {
			return err
		}
	}

	apiGroups, err := o.DiscoveryClient.ServerGroups()//获取apigroup
	if err != nil {
		return err
	}

	metricsAPIAvailable := SupportedMetricsAPIVersionAvailable(apiGroups)//判断Metrics是否支持这个apiGroup

	metrics := &metricsapi.NodeMetricsList{}
	if metricsAPIAvailable {//如果Metrics支持这个apiGroup
		metrics, err = getNodeMetricsFromMetricsAPI(o.MetricsClient, o.ResourceName, selector)//获取metrics
		if err != nil {
			return err
		}
	} else {//如果不支持
		metrics, err = o.Client.GetNodeMetrics(o.ResourceName, selector.String())//通过client获取metrics
		if err != nil {
			return err
		}
	}

	if len(metrics.Items) == 0 {//如果没有metrics
		return errors.New("metrics not available yet")
	}

	var nodes []v1.Node
	if len(o.ResourceName) > 0 {//如果资源有值
		node, err := o.NodeClient.Nodes().Get(o.ResourceName, metav1.GetOptions{})//获取资源名称对应节点
		if err != nil {
			return err
		}
		nodes = append(nodes, *node)//append节点
	} else {//如果selector有值
		nodeList, err := o.NodeClient.Nodes().List(metav1.ListOptions{//获取selector对应节点
			LabelSelector: selector.String(),
		})
		if err != nil {
			return err
		}
		nodes = append(nodes, nodeList.Items...)//append节点
	}

	allocatable := make(map[string]v1.ResourceList)//可分配资源map

	for _, n := range nodes {//遍历节点，设置可分配资源
		allocatable[n.Name] = n.Status.Allocatable
	}

	return o.Printer.PrintNodeMetrics(metrics.Items, allocatable, o.NoHeaders, o.SortBy)//打印结果
}

func getNodeMetricsFromMetricsAPI(metricsClient metricsclientset.Interface, resourceName string, selector labels.Selector) (*metricsapi.NodeMetricsList, error) {//获取metrics
	var err error
	versionedMetrics := &metricsV1beta1api.NodeMetricsList{}//构造metricsList对象
	mc := metricsClient.MetricsV1beta1()//获取client
	nm := mc.NodeMetricses()//获取node metrics client
	if resourceName != "" {//如果资源名称不为空
		m, err := nm.Get(resourceName, metav1.GetOptions{})//获取metrics
		if err != nil {
			return nil, err
		}
		versionedMetrics.Items = []metricsV1beta1api.NodeMetrics{*m}//包装metricsList
	} else {//如果selector有值
		versionedMetrics, err = nm.List(metav1.ListOptions{LabelSelector: selector.String()})//通过selector获取metrics
		if err != nil {
			return nil, err
		}
	}
	metrics := &metricsapi.NodeMetricsList{}
	err = metricsV1beta1api.Convert_v1beta1_NodeMetricsList_To_metrics_NodeMetricsList(versionedMetrics, metrics, nil)//把metrics转成NodeMetricsList
	if err != nil {
		return nil, err
	}
	return metrics, nil//返回NodeMetricsList
}

//打印node metrics
func (printer *TopCmdPrinter) PrintNodeMetrics(metrics []metricsapi.NodeMetrics, availableResources map[string]v1.ResourceList, noHeaders bool, sortBy string) error {
	if len(metrics) == 0 {//metrics为空返回
		return nil
	}
	w := printers.GetNewTabWriter(printer.out)//包装tab Writer
	defer w.Flush()//defer刷新

	n, err := NewNodeMetricsSorter(metrics, sortBy)//metrics排序
	if err != nil {
		return err
	}
	sort.Sort(n)//执行排序

	if !noHeaders {//如果不是noHeaders，打印headers
		printColumnNames(w, NodeColumns)
	}
	var usage v1.ResourceList
	for _, m := range metrics {//遍历metrics
		err := scheme.Scheme.Convert(&m.Usage, &usage, nil)//转换usage
		if err != nil {
			return err
		}
		printMetricsLine(w, &ResourceMetricsInfo{//打印一行metrics info
			Name:      m.Name,
			Metrics:   usage,
			Available: availableResources[m.Name],
		})
		delete(availableResources, m.Name)//删除availableResources 中已经打印的
	}

	// print lines for nodes of which the metrics is unreachable.
	for nodeName := range availableResources {//遍历availableResources，打印没有metrics的node
		printMissingMetricsNodeLine(w, nodeName)
	}
	return nil
}

//判断metrics是否支持apiGroup
func SupportedMetricsAPIVersionAvailable(discoveredAPIGroups *metav1.APIGroupList) bool {
	for _, discoveredAPIGroup := range discoveredAPIGroups.Groups {
		if discoveredAPIGroup.Name != metricsapi.GroupName {//如果groupname不同则跳过
			continue
		}
		for _, version := range discoveredAPIGroup.Versions {
			for _, supportedVersion := range supportedMetricsAPIVersions {
				if version.Version == supportedVersion {//如果version为supportedVersion 则返回true
					return true
				}
			}
		}
	}
	return false//返回false
}