Golang log.G函数代码示例

func (u *Updater) useExistingTask(ctx context.Context, slot slot, existing *api.Task) {
	var removeTasks []*api.Task
	for _, t := range slot {
		if t != existing {
			removeTasks = append(removeTasks, t)
		}
	}
	if len(removeTasks) != 0 || existing.DesiredState != api.TaskStateRunning {
		_, err := u.store.Batch(func(batch *store.Batch) error {
			u.removeOldTasks(ctx, batch, removeTasks)

			if existing.DesiredState != api.TaskStateRunning {
				err := batch.Update(func(tx store.Tx) error {
					t := store.GetTask(tx, existing.ID)
					if t == nil {
						return fmt.Errorf("task %s not found while trying to start it", existing.ID)
					}
					if t.DesiredState >= api.TaskStateRunning {
						return fmt.Errorf("task %s was already started when reached by updater", existing.ID)
					}
					t.DesiredState = api.TaskStateRunning
					return store.UpdateTask(tx, t)
				})
				if err != nil {
					log.G(ctx).WithError(err).Errorf("starting task %s failed", existing.ID)
				}
			}
			return nil
		})
		if err != nil {
			log.G(ctx).WithError(err).Error("updater batch transaction failed")
		}
	}
}

func deleteServiceTasks(ctx context.Context, s *store.MemoryStore, service *api.Service) {
	var (
		tasks []*api.Task
		err   error
	)
	s.View(func(tx store.ReadTx) {
		tasks, err = store.FindTasks(tx, store.ByServiceID(service.ID))
	})
	if err != nil {
		log.G(ctx).WithError(err).Errorf("failed to list tasks")
		return
	}

	_, err = s.Batch(func(batch *store.Batch) error {
		for _, t := range tasks {
			err := batch.Update(func(tx store.Tx) error {
				if err := store.DeleteTask(tx, t.ID); err != nil {
					log.G(ctx).WithError(err).Errorf("failed to delete task")
				}
				return nil
			})
			if err != nil {
				return err
			}
		}
		return nil
	})
	if err != nil {
		log.G(ctx).WithError(err).Errorf("task search transaction failed")
	}
}

func (u *Updater) rollbackUpdate(ctx context.Context, serviceID, message string) {
	log.G(ctx).Debugf("starting rollback of service %s", serviceID)

	var service *api.Service
	err := u.store.Update(func(tx store.Tx) error {
		service = store.GetService(tx, serviceID)
		if service == nil {
			return nil
		}
		if service.UpdateStatus == nil {
			// The service was updated since we started this update
			return nil
		}

		service.UpdateStatus.State = api.UpdateStatus_ROLLBACK_STARTED
		service.UpdateStatus.Message = message

		if service.PreviousSpec == nil {
			return errors.New("cannot roll back service because no previous spec is available")
		}
		service.Spec = *service.PreviousSpec
		service.PreviousSpec = nil

		return store.UpdateService(tx, service)
	})

	if err != nil {
		log.G(ctx).WithError(err).Errorf("failed to start rollback of service %s", serviceID)
		return
	}
}

func (g *GlobalOrchestrator) removeTasksFromNode(ctx context.Context, node *api.Node) {
	var (
		tasks []*api.Task
		err   error
	)
	g.store.View(func(tx store.ReadTx) {
		tasks, err = store.FindTasks(tx, store.ByNodeID(node.ID))
	})
	if err != nil {
		log.G(ctx).WithError(err).Errorf("global orchestrator: removeTasksFromNode failed finding tasks")
		return
	}

	_, err = g.store.Batch(func(batch *store.Batch) error {
		for _, t := range tasks {
			// GlobalOrchestrator only removes tasks from globalServices
			if _, exists := g.globalServices[t.ServiceID]; exists {
				g.removeTask(ctx, batch, t)
			}
		}
		return nil
	})
	if err != nil {
		log.G(ctx).WithError(err).Errorf("global orchestrator: removeTasksFromNode failed")
	}
}

func (na *NetworkAllocator) deallocateVIP(vip *api.Endpoint_VirtualIP) error {
	localNet := na.getNetwork(vip.NetworkID)
	if localNet == nil {
		return fmt.Errorf("networkallocator: could not find local network state")
	}

	ipam, _, err := na.resolveIPAM(localNet.nw)
	if err != nil {
		return fmt.Errorf("failed to resolve IPAM while allocating : %v", err)
	}

	// Retrieve the poolID and immediately nuke
	// out the mapping.
	poolID := localNet.endpoints[vip.Addr]
	delete(localNet.endpoints, vip.Addr)

	ip, _, err := net.ParseCIDR(vip.Addr)
	if err != nil {
		log.G(context.TODO()).Errorf("Could not parse VIP address %s while releasing", vip.Addr)
		return err
	}

	if err := ipam.ReleaseAddress(poolID, ip); err != nil {
		log.G(context.TODO()).Errorf("IPAM failure while releasing VIP address %s: %v", vip.Addr, err)
		return err
	}

	return nil
}

// getIDs returns an ordered set of IDs included in the given snapshot and
// the entries. The given snapshot/entries can contain two kinds of
// ID-related entry:
// - ConfChangeAddNode, in which case the contained ID will be added into the set.
// - ConfChangeRemoveNode, in which case the contained ID will be removed from the set.
func getIDs(snap *raftpb.Snapshot, ents []raftpb.Entry) []uint64 {
	ids := make(map[uint64]bool)
	if snap != nil {
		for _, id := range snap.Metadata.ConfState.Nodes {
			ids[id] = true
		}
	}
	for _, e := range ents {
		if e.Type != raftpb.EntryConfChange {
			continue
		}
		if snap != nil && e.Index < snap.Metadata.Index {
			continue
		}
		var cc raftpb.ConfChange
		if err := cc.Unmarshal(e.Data); err != nil {
			log.G(context.Background()).Panicf("unmarshal configuration change should never fail: %v", err)
		}
		switch cc.Type {
		case raftpb.ConfChangeAddNode:
			ids[cc.NodeID] = true
		case raftpb.ConfChangeRemoveNode:
			delete(ids, cc.NodeID)
		case raftpb.ConfChangeUpdateNode:
			// do nothing
		default:
			log.G(context.Background()).Panic("ConfChange Type should be either ConfChangeAddNode or ConfChangeRemoveNode!")
		}
	}
	var sids []uint64
	for id := range ids {
		sids = append(sids, id)
	}
	return sids
}

func (u *Updater) pauseUpdate(ctx context.Context, serviceID, message string) {
	log.G(ctx).Debugf("pausing update of service %s", serviceID)

	err := u.store.Update(func(tx store.Tx) error {
		service := store.GetService(tx, serviceID)
		if service == nil {
			return nil
		}
		if service.UpdateStatus == nil {
			// The service was updated since we started this update
			return nil
		}

		if service.UpdateStatus.State == api.UpdateStatus_ROLLBACK_STARTED {
			service.UpdateStatus.State = api.UpdateStatus_ROLLBACK_PAUSED
		} else {
			service.UpdateStatus.State = api.UpdateStatus_PAUSED
		}
		service.UpdateStatus.Message = message

		return store.UpdateService(tx, service)
	})

	if err != nil {
		log.G(ctx).WithError(err).Errorf("failed to pause update of service %s", serviceID)
	}
}

// start begins the session and returns the first SessionMessage.
func (s *session) start(ctx context.Context) error {
	log.G(ctx).Debugf("(*session).start")

	client := api.NewDispatcherClient(s.agent.config.Conn)

	description, err := s.agent.config.Executor.Describe(ctx)
	if err != nil {
		log.G(ctx).WithError(err).WithField("executor", s.agent.config.Executor).
			Errorf("node description unavailable")
		return err
	}
	// Override hostname
	if s.agent.config.Hostname != "" {
		description.Hostname = s.agent.config.Hostname
	}

	stream, err := client.Session(ctx, &api.SessionRequest{
		Description: description,
	})
	if err != nil {
		return err
	}

	msg, err := stream.Recv()
	if err != nil {
		return err
	}

	s.sessionID = msg.SessionID
	s.session = stream

	return s.handleSessionMessage(ctx, msg)
}

func (a *Allocator) procUnallocatedServices(ctx context.Context) {
	nc := a.netCtx
	var allocatedServices []*api.Service
	for _, s := range nc.unallocatedServices {
		if !nc.nwkAllocator.IsServiceAllocated(s) {
			if err := a.allocateService(ctx, s); err != nil {
				log.G(ctx).WithError(err).Debugf("Failed allocation of unallocated service %s", s.ID)
				continue
			}
			allocatedServices = append(allocatedServices, s)
		}
	}

	if len(allocatedServices) == 0 {
		return
	}

	committed, err := a.store.Batch(func(batch *store.Batch) error {
		for _, s := range allocatedServices {
			if err := a.commitAllocatedService(ctx, batch, s); err != nil {
				log.G(ctx).WithError(err).Debugf("Failed to commit allocation of unallocated service %s", s.ID)
				continue
			}
		}
		return nil
	})

	if err != nil {
		log.G(ctx).WithError(err).Error("Failed to commit allocation of unallocated services")
	}

	for _, s := range allocatedServices[:committed] {
		delete(nc.unallocatedServices, s.ID)
	}
}

// reconcileServiceOneNode checks one service on one node
func (g *GlobalOrchestrator) reconcileServiceOneNode(ctx context.Context, serviceID string, nodeID string) {
	_, exists := g.nodes[nodeID]
	if !exists {
		return
	}
	service, exists := g.globalServices[serviceID]
	if !exists {
		return
	}
	// the node has completed this servie
	completed := false
	// tasks for this node and service
	var (
		tasks []*api.Task
		err   error
	)
	g.store.View(func(tx store.ReadTx) {
		var tasksOnNode []*api.Task
		tasksOnNode, err = store.FindTasks(tx, store.ByNodeID(nodeID))
		if err != nil {
			return
		}
		for _, t := range tasksOnNode {
			// only interested in one service
			if t.ServiceID != serviceID {
				continue
			}
			if isTaskRunning(t) {
				tasks = append(tasks, t)
			} else {
				if isTaskCompleted(t, restartCondition(t)) {
					completed = true
				}
			}
		}
	})
	if err != nil {
		log.G(ctx).WithError(err).Errorf("global orchestrator: reconcile failed finding tasks")
		return
	}

	_, err = g.store.Batch(func(batch *store.Batch) error {
		// if restart policy considers this node has finished its task
		// it should remove all running tasks
		if completed {
			g.removeTasks(ctx, batch, service, tasks)
			return nil
		}
		// this node needs to run 1 copy of the task
		if len(tasks) == 0 {
			g.addTask(ctx, batch, service, nodeID)
		} else {
			g.removeTasks(ctx, batch, service, tasks[1:])
		}
		return nil
	})
	if err != nil {
		log.G(ctx).WithError(err).Errorf("global orchestrator: reconcileServiceOneNode batch failed")
	}
}

// updateCluster is called when there are cluster changes, and it ensures that the local RootCA is
// always aware of changes in clusterExpiry and the Root CA key material
func (s *Server) updateCluster(ctx context.Context, cluster *api.Cluster) {
	s.mu.Lock()
	s.joinTokens = cluster.RootCA.JoinTokens.Copy()
	s.mu.Unlock()
	var err error

	// If the cluster has a RootCA, let's try to update our SecurityConfig to reflect the latest values
	rCA := cluster.RootCA
	if len(rCA.CACert) != 0 && len(rCA.CAKey) != 0 {
		expiry := DefaultNodeCertExpiration
		if cluster.Spec.CAConfig.NodeCertExpiry != nil {
			// NodeCertExpiry exists, let's try to parse the duration out of it
			clusterExpiry, err := ptypes.Duration(cluster.Spec.CAConfig.NodeCertExpiry)
			if err != nil {
				log.G(ctx).WithFields(logrus.Fields{
					"cluster.id": cluster.ID,
					"method":     "(*Server).updateCluster",
				}).WithError(err).Warn("failed to parse certificate expiration, using default")
			} else {
				// We were able to successfully parse the expiration out of the cluster.
				expiry = clusterExpiry
			}
		} else {
			// NodeCertExpiry seems to be nil
			log.G(ctx).WithFields(logrus.Fields{
				"cluster.id": cluster.ID,
				"method":     "(*Server).updateCluster",
			}).WithError(err).Warn("failed to parse certificate expiration, using default")

		}
		// Attempt to update our local RootCA with the new parameters
		err = s.securityConfig.UpdateRootCA(rCA.CACert, rCA.CAKey, expiry)
		if err != nil {
			log.G(ctx).WithFields(logrus.Fields{
				"cluster.id": cluster.ID,
				"method":     "(*Server).updateCluster",
			}).WithError(err).Error("updating Root CA failed")
		} else {
			log.G(ctx).WithFields(logrus.Fields{
				"cluster.id": cluster.ID,
				"method":     "(*Server).updateCluster",
			}).Debugf("Root CA updated successfully")
		}
	}

	// Update our security config with the list of External CA URLs
	// from the new cluster state.

	// TODO(aaronl): In the future, this will be abstracted with an
	// ExternalCA interface that has different implementations for
	// different CA types. At the moment, only CFSSL is supported.
	var cfsslURLs []string
	for _, ca := range cluster.Spec.CAConfig.ExternalCAs {
		if ca.Protocol == api.ExternalCA_CAProtocolCFSSL {
			cfsslURLs = append(cfsslURLs, ca.URL)
		}
	}

	s.securityConfig.externalCA.UpdateURLs(cfsslURLs...)
}

// Init prepares the worker for assignments.
func (w *worker) Init(ctx context.Context) error {
	w.mu.Lock()
	defer w.mu.Unlock()

	ctx = log.WithModule(ctx, "worker")

	// TODO(stevvooe): Start task cleanup process.

	// read the tasks from the database and start any task managers that may be needed.
	return w.db.Update(func(tx *bolt.Tx) error {
		return WalkTasks(tx, func(task *api.Task) error {
			if !TaskAssigned(tx, task.ID) {
				// NOTE(stevvooe): If tasks can survive worker restart, we need
				// to startup the controller and ensure they are removed. For
				// now, we can simply remove them from the database.
				if err := DeleteTask(tx, task.ID); err != nil {
					log.G(ctx).WithError(err).Errorf("error removing task %v", task.ID)
				}
				return nil
			}

			status, err := GetTaskStatus(tx, task.ID)
			if err != nil {
				log.G(ctx).WithError(err).Error("unable to read tasks status")
				return nil
			}

			task.Status = *status // merges the status into the task, ensuring we start at the right point.
			return w.startTask(ctx, tx, task)
		})
	})
}

func (u *Updater) worker(ctx context.Context, queue <-chan *api.Task) {
	for t := range queue {
		updated := newTask(u.cluster, u.newService, t.Slot)
		updated.DesiredState = api.TaskStateReady
		if isGlobalService(u.newService) {
			updated.NodeID = t.NodeID
		}

		if err := u.updateTask(ctx, t, updated); err != nil {
			log.G(ctx).WithError(err).WithField("task.id", t.ID).Error("update failed")
		}

		if u.newService.Spec.Update != nil && (u.newService.Spec.Update.Delay.Seconds != 0 || u.newService.Spec.Update.Delay.Nanos != 0) {
			delay, err := ptypes.Duration(&u.newService.Spec.Update.Delay)
			if err != nil {
				log.G(ctx).WithError(err).Error("invalid update delay")
				continue
			}
			select {
			case <-time.After(delay):
			case <-u.stopChan:
				return
			}
		}
	}
}

// events issues a call to the events API and returns a channel with all
// events. The stream of events can be shutdown by cancelling the context.
func (c *containerAdapter) events(ctx context.Context) <-chan events.Message {
	log.G(ctx).Debugf("waiting on events")
	buffer, l := c.backend.SubscribeToEvents(time.Time{}, time.Time{}, c.container.eventFilter())
	eventsq := make(chan events.Message, len(buffer))

	for _, event := range buffer {
		eventsq <- event
	}

	go func() {
		defer c.backend.UnsubscribeFromEvents(l)

		for {
			select {
			case ev := <-l:
				jev, ok := ev.(events.Message)
				if !ok {
					log.G(ctx).Warnf("unexpected event message: %q", ev)
					continue
				}
				select {
				case eventsq <- jev:
				case <-ctx.Done():
					return
				}
			case <-ctx.Done():
				return
			}
		}
	}()

	return eventsq
}

// UpdateTaskStatus attempts to send a task status update over the current session,
// blocking until the operation is completed.
//
// If an error is returned, the operation should be retried.
func (a *Agent) UpdateTaskStatus(ctx context.Context, taskID string, status *api.TaskStatus) error {
	log.G(ctx).WithField("task.id", taskID).Debugf("(*Agent).UpdateTaskStatus")
	ctx, cancel := context.WithCancel(ctx)
	defer cancel()

	errs := make(chan error, 1)
	if err := a.withSession(ctx, func(session *session) error {
		go func() {
			err := session.sendTaskStatus(ctx, taskID, status)
			if err != nil {
				if err == errTaskUnknown {
					err = nil // dispatcher no longer cares about this task.
				} else {
					log.G(ctx).WithError(err).Error("sending task status update failed")
				}
			} else {
				log.G(ctx).Debug("task status reported")
			}

			errs <- err
		}()

		return nil
	}); err != nil {
		return err
	}

	select {
	case err := <-errs:
		return err
	case <-ctx.Done():
		return ctx.Err()
	}
}