Docker Swarm Code Analysis Notes (9)--swarm Cluster,engine and Addengine

This is a creation in Article, where the information may have evolved or changed.

swarm/cluster.goBelonging swarm package to this, it defines swarm the driver structure of the Cluster body:

// Cluster is exportedtype Cluster struct {    sync.RWMutex    eventHandlers     *cluster.EventHandlers    engines           map[string]*cluster.Engine    pendingEngines    map[string]*cluster.Engine    scheduler         *scheduler.Scheduler    discovery         discovery.Backend    pendingContainers map[string]*pendingContainer    overcommitRatio float64    engineOpts      *cluster.EngineOpts    createRetry     int64    TLSConfig       *tls.Config}

It is important that the cluster.Engine definition of this struct ( cluster/engine.go ), each cluster.Engine representing one Docker engine , namely Docker daemon :

// Engine represents a docker enginetype Engine struct {    sync.RWMutex    ID  string    IP  string    Addr string    Name string    Cpus int64    Memory  int64    Labels  map[string]string    Version string    stopCh  chan struct{}    refreshDelayer  *delayer    containers  map[string]*Container    images  []*Image    networksmap[string]*Network    volumes map[string]*Volume    client  dockerclient.Client    apiClient   engineapi.APIClient    eventHandlerEventHandler    state   engineState    lastError   string    updatedAt   time.Time    failureCountint    overcommitRatio int64    opts*EngineOpts    eventsMonitor   *EventsMonitor}

The cluster most important part of creating addEngine this method is Docker engine adding to cluster the medium:

  func (c *cluster) addengine (addr string) bool {//Check the engine is already registered by address. If C.hasenginebyaddr (addr) {return false} Engine: = cluster. Newengine (addr, c.overcommitratio, c.engineopts) if err: = engine. Registereventhandler (c); Err! = Nil {log. Error (ERR)}//ADD it to Pending engine map, indexed by address. This would prevent//duplicates from entering C.lock () c.pendingengines[addr] = Engine C.unlock ()//Valida    Tependingengine'll start a thread to validate the engine.    If the engine is reachable and valid, it's ll be monitored and updated in a loop. If engine is not reachable, pending engines'll be examined once in a while Go C.validatependingengine (engine) R Eturn true}  

addEngineFirst check whether the current is engine already present cluster , if not, it will be assigned a new one engine , add it first Cluster.pendingEngines , and start a new goroutine ( validatePendingEngine ) to check if it is a valid one engine .

In addition, the Engine.RegisterEventHandler member is essentially Cluster eventHandlers assigned to Engine.eventHandler :

// RegisterEventHandler registers an event handler.func (e *Engine) RegisterEventHandler(h EventHandler) error {    if e.eventHandler != nil {        return errors.New("event handler already set")    }    e.eventHandler = h    return nil}

It Cluster is eventHandlers assigned in newprimary, so it is essentially a set of Engine Cluster processing functions.