想系统学习k8s源码，云原生的可以加：mkjnnm

代码入口

controller-manager 代码入口如下图：

NewControllerManagerCommand

Controller Manager 也是一个命令行，通过一系列 flag 启动，具体的各个flag 我们就不多看，有兴趣的可以去文档或者 cmd/kube-controller-manager/app/options/options.go 文件里面去过滤一下，我们直接从Run 函数入手。

// NewControllerManagerCommand creates a *cobra.Command object with default parameters
// 定义了一个名为NewControllerManagerCommand的函数，它返回一个*cobra.Command类型的指针。Cobra是一个用于创建现代CLI应用程序的Go库。
func NewControllerManagerCommand() *cobra.Command {
	s, err := options.NewKubeControllerManagerOptions()
	if err != nil {
		klog.Background().Error(err, "Unable to initialize command options")
		klog.FlushAndExit(klog.ExitFlushTimeout, 1)
	}

	cmd := &cobra.Command{
		Use: "kube-controller-manager",
		Long: `The Kubernetes controller manager is a daemon that embeds
the core control loops shipped with Kubernetes. In applications of robotics and
automation, a control loop is a non-terminating loop that regulates the state of
the system. In Kubernetes, a controller is a control loop that watches the shared
state of the cluster through the apiserver and makes changes attempting to move the
current state towards the desired state. Examples of controllers that ship with
Kubernetes today are the replication controller, endpoints controller, namespace
controller, and serviceaccounts controller.`,
		PersistentPreRunE: func(*cobra.Command, []string) error {
			// silence client-go warnings.
			// kube-controller-manager generically watches APIs (including deprecated ones),
			// and CI ensures it works properly against matching kube-apiserver versions.
			// 在执行命令之前运行的函数，这里用于设置默认的警告处理器，以避免显示client-go的警告信息。
			restclient.SetDefaultWarningHandler(restclient.NoWarnings{})
			return nil
		},
		RunE: func(cmd *cobra.Command, args []string) error {
			verflag.PrintAndExitIfRequested()

			// Activate logging as soon as possible, after that
			// show flags with the final logging configuration.
			if err := logsapi.ValidateAndApply(s.Logs, utilfeature.DefaultFeatureGate); err != nil {
				return err
			}
			cliflag.PrintFlags(cmd.Flags())

			c, err := s.Config(KnownControllers(), ControllersDisabledByDefault(), ControllerAliases())
			if err != nil {
				return err
			}
			// add feature enablement metrics
			utilfeature.DefaultMutableFeatureGate.AddMetrics()
			return Run(context.Background(), c.Complete())
		},
		Args: func(cmd *cobra.Command, args []string) error {
			for _, arg := range args {
				if len(arg) > 0 {
					return fmt.Errorf("%q does not take any arguments, got %q", cmd.CommandPath(), args)
				}
			}
			return nil
		},
	}

	fs := cmd.Flags()
	namedFlagSets := s.Flags(KnownControllers(), ControllersDisabledByDefault(), ControllerAliases())
	verflag.AddFlags(namedFlagSets.FlagSet("global"))
	globalflag.AddGlobalFlags(namedFlagSets.FlagSet("global"), cmd.Name(), logs.SkipLoggingConfigurationFlags())
	for _, f := range namedFlagSets.FlagSets {
		fs.AddFlagSet(f)
	}

	cols, _, _ := term.TerminalSize(cmd.OutOrStdout())
	cliflag.SetUsageAndHelpFunc(cmd, namedFlagSets, cols)

	return cmd
}

Run Function 启动流程

Kube Controller Manager 既可以单实例启动，也可以多实例启动。 如果为了保证 HA 而启动多个Controller Manager，它就需要选主来保证同一时间只有一个Master 实例。我们来看一眼Run 函数的启动流程，这里粗看上去都是在选主:

// Run runs the KubeControllerManagerOptions.
func Run(ctx context.Context, c *config.CompletedConfig) error {
	logger := klog.FromContext(ctx)
	stopCh := ctx.Done()

	// To help debugging, immediately log version
	logger.Info("Starting", "version", version.Get())

	logger.Info("Golang settings", "GOGC", os.Getenv("GOGC"), "GOMAXPROCS", os.Getenv("GOMAXPROCS"), "GOTRACEBACK", os.Getenv("GOTRACEBACK"))

	// Start events processing pipeline.
	c.EventBroadcaster.StartStructuredLogging(0)
	c.EventBroadcaster.StartRecordingToSink(&v1core.EventSinkImpl{Interface: c.Client.CoreV1().Events("")})
	defer c.EventBroadcaster.Shutdown()

	if cfgz, err := configz.New(ConfigzName); err == nil {
		cfgz.Set(c.ComponentConfig)
	} else {
		logger.Error(err, "Unable to register configz")
	}

	// Setup any healthz checks we will want to use.
	var checks []healthz.HealthChecker
	var electionChecker *leaderelection.HealthzAdaptor
	if c.ComponentConfig.Generic.LeaderElection.LeaderElect {
		electionChecker = leaderelection.NewLeaderHealthzAdaptor(time.Second * 20)
		checks = append(checks, electionChecker)
	}
	healthzHandler := controllerhealthz.NewMutableHealthzHandler(checks...)

	// Start the controller manager HTTP server
	// unsecuredMux is the handler for these controller *after* authn/authz filters have been applied
	var unsecuredMux *mux.PathRecorderMux
	if c.SecureServing != nil {
		unsecuredMux = genericcontrollermanager.NewBaseHandler(&c.ComponentConfig.Generic.Debugging, healthzHandler)
		slis.SLIMetricsWithReset{}.Install(unsecuredMux)

		handler := genericcontrollermanager.BuildHandlerChain(unsecuredMux, &c.Authorization, &c.Authentication)
		// TODO: handle stoppedCh and listenerStoppedCh returned by c.SecureServing.Serve
		if _, _, err := c.SecureServing.Serve(handler, 0, stopCh); err != nil {
			return err
		}
	}

	clientBuilder, rootClientBuilder := createClientBuilders(logger, c)

	saTokenControllerDescriptor := newServiceAccountTokenControllerDescriptor(rootClientBuilder)

	run := func(ctx context.Context, controllerDescriptors map[string]*ControllerDescriptor) {
		controllerContext, err := CreateControllerContext(ctx, c, rootClientBuilder, clientBuilder)
		if err != nil {
			logger.Error(err, "Error building controller context")
			klog.FlushAndExit(klog.ExitFlushTimeout, 1)
		}

		if err := StartControllers(ctx, controllerContext, controllerDescriptors, unsecuredMux, healthzHandler); err != nil {
			logger.Error(err, "Error starting controllers")
			klog.FlushAndExit(klog.ExitFlushTimeout, 1)
		}

		controllerContext.InformerFactory.Start(stopCh)
		controllerContext.ObjectOrMetadataInformerFactory.Start(stopCh)
		close(controllerContext.InformersStarted)

		<-ctx.Done()
	}

	// No leader election, run directly
	if !c.ComponentConfig.Generic.LeaderElection.LeaderElect {
		controllerDescriptors := NewControllerDescriptors()
		controllerDescriptors[names.ServiceAccountTokenController] = saTokenControllerDescriptor
		run(ctx, controllerDescriptors)
		return nil
	}

	id, err := os.Hostname()
	if err != nil {
		return err
	}

	// add a uniquifier so that two processes on the same host don't accidentally both become active
	id = id + "_" + string(uuid.NewUUID())

	// leaderMigrator will be non-nil if and only if Leader Migration is enabled.
	var leaderMigrator *leadermigration.LeaderMigrator = nil

	// If leader migration is enabled, create the LeaderMigrator and prepare for migration
	if leadermigration.Enabled(&c.ComponentConfig.Generic) {
		logger.Info("starting leader migration")

		leaderMigrator = leadermigration.NewLeaderMigrator(&c.ComponentConfig.Generic.LeaderMigration,
			"kube-controller-manager")

		// startSATokenControllerInit is the original InitFunc.
		startSATokenControllerInit := saTokenControllerDescriptor.GetInitFunc()

		// Wrap saTokenControllerDescriptor to signal readiness for migration after starting
		//  the controller.
		saTokenControllerDescriptor.initFunc = func(ctx context.Context, controllerContext ControllerContext, controllerName string) (controller.Interface, bool, error) {
			defer close(leaderMigrator.MigrationReady)
			return startSATokenControllerInit(ctx, controllerContext, controllerName)
		}
	}

	// Start the main lock
	go leaderElectAndRun(ctx, c, id, electionChecker,
		c.ComponentConfig.Generic.LeaderElection.ResourceLock,
		c.ComponentConfig.Generic.LeaderElection.ResourceName,
		leaderelection.LeaderCallbacks{
			OnStartedLeading: func(ctx context.Context) {
				controllerDescriptors := NewControllerDescriptors()
				if leaderMigrator != nil {
					// If leader migration is enabled, we should start only non-migrated controllers
					//  for the main lock.
					controllerDescriptors = filteredControllerDescriptors(controllerDescriptors, leaderMigrator.FilterFunc, leadermigration.ControllerNonMigrated)
					logger.Info("leader migration: starting main controllers.")
				}
				controllerDescriptors[names.ServiceAccountTokenController] = saTokenControllerDescriptor
				run(ctx, controllerDescriptors)
			},
			OnStoppedLeading: func() {
				logger.Error(nil, "leaderelection lost")
				klog.FlushAndExit(klog.ExitFlushTimeout, 1)
			},
		})

	// If Leader Migration is enabled, proceed to attempt the migration lock.
	if leaderMigrator != nil {
		// Wait for Service Account Token Controller to start before acquiring the migration lock.
		// At this point, the main lock must have already been acquired, or the KCM process already exited.
		// We wait for the main lock before acquiring the migration lock to prevent the situation
		//  where KCM instance A holds the main lock while KCM instance B holds the migration lock.
		<-leaderMigrator.MigrationReady

		// Start the migration lock.
		go leaderElectAndRun(ctx, c, id, electionChecker,
			c.ComponentConfig.Generic.LeaderMigration.ResourceLock,
			c.ComponentConfig.Generic.LeaderMigration.LeaderName,
			leaderelection.LeaderCallbacks{
				OnStartedLeading: func(ctx context.Context) {
					logger.Info("leader migration: starting migrated controllers.")
					controllerDescriptors := NewControllerDescriptors()
					controllerDescriptors = filteredControllerDescriptors(controllerDescriptors, leaderMigrator.FilterFunc, leadermigration.ControllerMigrated)
					// DO NOT start saTokenController under migration lock
					delete(controllerDescriptors, names.ServiceAccountTokenController)
					run(ctx, controllerDescriptors)
				},
				OnStoppedLeading: func() {
					logger.Error(nil, "migration leaderelection lost")
					klog.FlushAndExit(klog.ExitFlushTimeout, 1)
				},
			})
	}

	<-stopCh
	return nil
}

Run

如果选主成功，会执行 run方法：

	controllerDescriptors := NewControllerDescriptors()
				
	run := func(ctx context.Context, controllerDescriptors map[string]*ControllerDescriptor) {
		controllerContext, err := CreateControllerContext(ctx, c, rootClientBuilder, clientBuilder)
		if err != nil {
			logger.Error(err, "Error building controller context")
			klog.FlushAndExit(klog.ExitFlushTimeout, 1)
		}
		
		// Run 方法中，主要函数是  StartControllers
		if err := StartControllers(ctx, controllerContext, controllerDescriptors, unsecuredMux, healthzHandler); err != nil {
			logger.Error(err, "Error starting controllers")
			klog.FlushAndExit(klog.ExitFlushTimeout, 1)
		}

		controllerContext.InformerFactory.Start(stopCh)
		controllerContext.ObjectOrMetadataInformerFactory.Start(stopCh)
		close(controllerContext.InformersStarted)

		<-ctx.Done()
	}

StartControllers

首先启动 SA token controller ，因为其他控制器需要依赖他，然后 for 循环遍历 controllerDescriptor, 调用 startController 方法

// StartControllers starts a set of controllers with a specified ControllerContext
func StartControllers(ctx context.Context, controllerCtx ControllerContext, controllerDescriptors map[string]*ControllerDescriptor,
	unsecuredMux *mux.PathRecorderMux, healthzHandler *controllerhealthz.MutableHealthzHandler) error {
	var controllerChecks []healthz.HealthChecker

	// Always start the SA token controller first using a full-power client, since it needs to mint tokens for the rest
	// If this fails, just return here and fail since other controllers won't be able to get credentials.
	if serviceAccountTokenControllerDescriptor, ok := controllerDescriptors[names.ServiceAccountTokenController]; ok {
		check, err := StartController(ctx, controllerCtx, serviceAccountTokenControllerDescriptor, unsecuredMux)
		if err != nil {
			return err
		}
		if check != nil {
			// HealthChecker should be present when controller has started
			controllerChecks = append(controllerChecks, check)
		}
	}

	// Initialize the cloud provider with a reference to the clientBuilder only after token controller
	// has started in case the cloud provider uses the client builder.
	if controllerCtx.Cloud != nil {
		controllerCtx.Cloud.Initialize(controllerCtx.ClientBuilder, ctx.Done())
	}

	// Each controller is passed a context where the logger has the name of
	// the controller set through WithName. That name then becomes the prefix of
	// of all log messages emitted by that controller.
	//
	// In StartController, an explicit "controller" key is used instead, for two reasons:
	// - while contextual logging is alpha, klog.LoggerWithName is still a no-op,
	//   so we cannot rely on it yet to add the name
	// - it allows distinguishing between log entries emitted by the controller
	//   and those emitted for it - this is a bit debatable and could be revised.
	for _, controllerDesc := range controllerDescriptors {
		if controllerDesc.RequiresSpecialHandling() {
			continue
		}
		
		check, err := StartController(ctx, controllerCtx, controllerDesc, unsecuredMux)
		if err != nil {
			return err
		}
		if check != nil {
			// HealthChecker should be present when controller has started
			controllerChecks = append(controllerChecks, check)
		}
	}

	healthzHandler.AddHealthChecker(controllerChecks...)

	return nil
}

StartContrller

// StartController starts a controller with a specified ControllerContext
// and performs required pre- and post- checks/actions
func StartController(ctx context.Context, controllerCtx ControllerContext, controllerDescriptor *ControllerDescriptor,
	unsecuredMux *mux.PathRecorderMux) (healthz.HealthChecker, error) {
	logger := klog.FromContext(ctx)
	controllerName := controllerDescriptor.Name()

	for _, featureGate := range controllerDescriptor.GetRequiredFeatureGates() {
		if !utilfeature.DefaultFeatureGate.Enabled(featureGate) {
			logger.Info("Controller is disabled by a feature gate", "controller", controllerName, "requiredFeatureGates", controllerDescriptor.GetRequiredFeatureGates())
			return nil, nil
		}
	}

	if controllerDescriptor.IsCloudProviderController() && controllerCtx.LoopMode != IncludeCloudLoops {
		logger.Info("Skipping a cloud provider controller", "controller", controllerName, "loopMode", controllerCtx.LoopMode)
		return nil, nil
	}

	if !controllerCtx.IsControllerEnabled(controllerDescriptor) {
		logger.Info("Warning: controller is disabled", "controller", controllerName)
		return nil, nil
	}

	time.Sleep(wait.Jitter(controllerCtx.ComponentConfig.Generic.ControllerStartInterval.Duration, ControllerStartJitter))

	logger.V(1).Info("Starting controller", "controller", controllerName)

	// 获取 controllerDescriptor 的 InitFunc 并调用
	initFunc := controllerDescriptor.GetInitFunc()
	ctrl, started, err := initFunc(klog.NewContext(ctx, klog.LoggerWithName(logger, controllerName)), controllerCtx, controllerName)
	if err != nil {
		logger.Error(err, "Error starting controller", "controller", controllerName)
		return nil, err
	}
	if !started {
		logger.Info("Warning: skipping controller", "controller", controllerName)
		return nil, nil
	}

	check := controllerhealthz.NamedPingChecker(controllerName)
	if ctrl != nil {
		// check if the controller supports and requests a debugHandler
		// and it needs the unsecuredMux to mount the handler onto.
		if debuggable, ok := ctrl.(controller.Debuggable); ok && unsecuredMux != nil {
			if debugHandler := debuggable.DebuggingHandler(); debugHandler != nil {
				basePath := "/debug/controllers/" + controllerName
				unsecuredMux.UnlistedHandle(basePath, http.StripPrefix(basePath, debugHandler))
				unsecuredMux.UnlistedHandlePrefix(basePath+"/", http.StripPrefix(basePath, debugHandler))
			}
		}
		if healthCheckable, ok := ctrl.(controller.HealthCheckable); ok {
			if realCheck := healthCheckable.HealthChecker(); realCheck != nil {
				check = controllerhealthz.NamedHealthChecker(controllerName, realCheck)
			}
		}
	}

	logger.Info("Started controller", "controller", controllerName)
	return check, nil
}

InitFunc

所有的控制器都会将其启动方法包装成 inifFunc，这里我们来看下 deployment , 代码在 cmd/kube-controller-manager/app/controllermanager.go

func newDeploymentControllerDescriptor() *ControllerDescriptor {
	return &ControllerDescriptor{
		name:     names.DeploymentController,
		aliases:  []string{"deployment"},
		initFunc: startDeploymentController,
	}
}

func startDeploymentController(ctx context.Context, controllerContext ControllerContext, controllerName string) (controller.Interface, bool, error) {
	dc, err := deployment.NewDeploymentController(
		ctx,
		controllerContext.InformerFactory.Apps().V1().Deployments(),
		controllerContext.InformerFactory.Apps().V1().ReplicaSets(),
		controllerContext.InformerFactory.Core().V1().Pods(),
		controllerContext.ClientBuilder.ClientOrDie("deployment-controller"),
	)
	if err != nil {
		return nil, true, fmt.Errorf("error creating Deployment controller: %v", err)
	}
	go dc.Run(ctx, int(controllerContext.ComponentConfig.DeploymentController.ConcurrentDeploymentSyncs))
	return nil, true, nil
}