Merge pull request #37484 from thaJeztah/bump_swarmkit

Update swarmkit to 68266392a176434d282760d2d6d0ab4c68edcae6

vendor.conf

@@ -125,7 +125,7 @@ github.com/containerd/ttrpc 94dde388801693c54f88a6596f713b51a8b30b2d
 github.com/gogo/googleapis 08a7655d27152912db7aaf4f983275eaf8d128ef
 
 # cluster
-github.com/docker/swarmkit 199cf49cd99690135d99e52a1907ec82e8113c4f
+github.com/docker/swarmkit 68266392a176434d282760d2d6d0ab4c68edcae6
 github.com/gogo/protobuf v1.0.0
 github.com/cloudflare/cfssl 1.3.2
 github.com/fernet/fernet-go 1b2437bc582b3cfbb341ee5a29f8ef5b42912ff2

vendor/github.com/docker/swarmkit/manager/dispatcher/dispatcher.go

@@ -293,9 +293,16 @@ func (d *Dispatcher) Run(ctx context.Context) error {
 			publishManagers(ev.([]*api.Peer))
 		case <-d.processUpdatesTrigger:
 			d.processUpdates(ctx)
+			batchTimer.Stop()
+			// drain the timer, if it has already expired
+			select {
+			case <-batchTimer.C:
+			default:
+			}
 			batchTimer.Reset(maxBatchInterval)
 		case <-batchTimer.C:
 			d.processUpdates(ctx)
+			// batch timer has already expired, so no need to drain
 			batchTimer.Reset(maxBatchInterval)
 		case v := <-configWatcher:
 			cluster := v.(api.EventUpdateCluster)
@@ -416,7 +423,7 @@ func (d *Dispatcher) markNodesUnknown(ctx context.Context) error {
 
 				expireFunc := func() {
 					log := log.WithField("node", nodeID)
-					log.Info(`heartbeat expiration for node %s in state "unknown"`, nodeID)
+					log.Infof(`heartbeat expiration for node %s in state "unknown"`, nodeID)
 					if err := d.markNodeNotReady(nodeID, api.NodeStatus_DOWN, `heartbeat failure for node in "unknown" state`); err != nil {
 						log.WithError(err).Error(`failed deregistering node after heartbeat expiration for node in "unknown" state`)
 					}
@@ -537,7 +544,7 @@ func (d *Dispatcher) register(ctx context.Context, nodeID string, description *a
 	}
 
 	expireFunc := func() {
-		log.G(ctx).Debug("heartbeat expiration for worker %s, setting worker status to NodeStatus_DOWN ", nodeID)
+		log.G(ctx).Debugf("heartbeat expiration for worker %s, setting worker status to NodeStatus_DOWN ", nodeID)
 		if err := d.markNodeNotReady(nodeID, api.NodeStatus_DOWN, "heartbeat failure"); err != nil {
 			log.G(ctx).WithError(err).Errorf("failed deregistering node after heartbeat expiration")
 		}

vendor/github.com/docker/swarmkit/manager/scheduler/scheduler.go

@@ -702,9 +702,20 @@ func (s *Scheduler) scheduleNTasksOnNodes(ctx context.Context, n int, taskGroup
 	return tasksScheduled
 }
 
+// noSuitableNode checks unassigned tasks and make sure they have an existing service in the store before
+// updating the task status and adding it back to: schedulingDecisions, unassignedTasks and allTasks
 func (s *Scheduler) noSuitableNode(ctx context.Context, taskGroup map[string]*api.Task, schedulingDecisions map[string]schedulingDecision) {
 	explanation := s.pipeline.Explain()
 	for _, t := range taskGroup {
+		var service *api.Service
+		s.store.View(func(tx store.ReadTx) {
+			service = store.GetService(tx, t.ServiceID)
+		})
+		if service == nil {
+			log.G(ctx).WithField("task.id", t.ID).Debug("removing task from the scheduler")
+			continue
+		}
+
 		log.G(ctx).WithField("task.id", t.ID).Debug("no suitable node available for task")
 
 		newT := *t

vendor/github.com/docker/swarmkit/node/node.go

@@ -1136,6 +1136,11 @@ func (n *Node) superviseManager(ctx context.Context, securityConfig *ca.Security
 			// re-promoted. In this case, we must assume we were
 			// re-promoted, and restart the manager.
 			log.G(ctx).Warn("failed to get worker role after manager stop, forcing certificate renewal")
+
+			// We can safely reset this timer without stopping/draining the timer
+			// first because the only way the code has reached this point is if the timer
+			// has already expired - if the role changed or the context were canceled,
+			// then we would have returned already.
 			timer.Reset(roleChangeTimeout)
 
 			renewer.Renew()