Merge pull request #353 from thaJeztah/19.03_bump_swarmkit

[19.03 backport] bump swarmkit to bbe341867eae1615faf8a702ec05bfe986e73e06 (bump_v19.03 branch)

vendor.conf

@@ -130,7 +130,7 @@ github.com/containerd/ttrpc                         699c4e40d1e7416e08bf7019c7ce
 github.com/gogo/googleapis                          d31c731455cb061f42baff3bda55bad0118b126b # v1.2.0
 
 # cluster
-github.com/docker/swarmkit                          4fb9e961aba635f0240a140e89ece6d6c2082585 # bump_v19.03 branch
+github.com/docker/swarmkit                          bbe341867eae1615faf8a702ec05bfe986e73e06 # bump_v19.03 branch
 github.com/gogo/protobuf                            ba06b47c162d49f2af050fb4c75bcbc86a159d5c # v1.2.1
 github.com/cloudflare/cfssl                         5d63dbd981b5c408effbb58c442d54761ff94fbd # 1.3.2
 github.com/fernet/fernet-go                         1b2437bc582b3cfbb341ee5a29f8ef5b42912ff2

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

@@ -575,7 +575,7 @@ func (a *Agent) nodeDescriptionWithHostname(ctx context.Context, tlsInfo *api.No
 
 	// Override hostname and TLS info
 	if desc != nil {
-		if a.config.Hostname != "" && desc != nil {
+		if a.config.Hostname != "" {
 			desc.Hostname = a.config.Hostname
 		}
 		desc.TLSInfo = tlsInfo

vendor/github.com/docker/swarmkit/agent/session.go

@@ -3,6 +3,7 @@ package agent
 import (
 	"context"
 	"errors"
+	"math"
 	"sync"
 	"time"
 
@@ -64,6 +65,7 @@ func newSession(ctx context.Context, agent *Agent, delay time.Duration, sessionI
 	cc, err := agent.config.ConnBroker.Select(
 		grpc.WithTransportCredentials(agent.config.Credentials),
 		grpc.WithTimeout(dispatcherRPCTimeout),
+		grpc.WithDefaultCallOptions(grpc.MaxCallRecvMsgSize(math.MaxInt32)),
 	)
 
 	if err != nil {
@@ -136,7 +138,7 @@ func (s *session) start(ctx context.Context, description *api.NodeDescription) e
 	// `ctx` is done and hence fail to propagate the timeout error to the agent.
 	// If the error is not propogated to the agent, the agent will not close
 	// the session or rebuild a new session.
-	sessionCtx, cancelSession := context.WithCancel(ctx) // nolint: vet
+	sessionCtx, cancelSession := context.WithCancel(ctx) //nolint:govet
 
 	// Need to run Session in a goroutine since there's no way to set a
 	// timeout for an individual Recv call in a stream.
@@ -159,7 +161,7 @@ func (s *session) start(ctx context.Context, description *api.NodeDescription) e
 	select {
 	case err := <-errChan:
 		if err != nil {
-			return err // nolint: vet
+			return err //nolint:govet
 		}
 	case <-time.After(dispatcherRPCTimeout):
 		cancelSession()

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

@@ -238,7 +238,7 @@ func (d *Dispatcher) Run(ctx context.Context) error {
 			if err != nil {
 				return err
 			}
-			if err == nil && len(clusters) == 1 {
+			if len(clusters) == 1 {
 				heartbeatPeriod, err := gogotypes.DurationFromProto(clusters[0].Spec.Dispatcher.HeartbeatPeriod)
 				if err == nil && heartbeatPeriod > 0 {
 					d.config.HeartbeatPeriod = heartbeatPeriod

vendor/github.com/docker/swarmkit/manager/drivers/provider.go

@@ -22,7 +22,7 @@ func (m *DriverProvider) NewSecretDriver(driver *api.Driver) (*SecretDriver, err
 	if m.pluginGetter == nil {
 		return nil, fmt.Errorf("plugin getter is nil")
 	}
-	if driver == nil && driver.Name == "" {
+	if driver == nil || driver.Name == "" {
 		return nil, fmt.Errorf("driver specification is nil")
 	}
 	// Search for the specified plugin

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

@@ -4,6 +4,7 @@ import (
 	"context"
 	"crypto/tls"
 	"fmt"
+	"math"
 	"net"
 	"os"
 	"path/filepath"
@@ -758,6 +759,7 @@ func (m *Manager) updateKEK(ctx context.Context, cluster *api.Cluster) error {
 					func(addr string, timeout time.Duration) (net.Conn, error) {
 						return xnet.DialTimeoutLocal(addr, timeout)
 					}),
+				grpc.WithDefaultCallOptions(grpc.MaxCallRecvMsgSize(math.MaxInt32)),
 			)
 			if err != nil {
 				logger.WithError(err).Error("failed to connect to local manager socket after locking the cluster")

vendor/github.com/docker/swarmkit/manager/orchestrator/restart/restart.go

@@ -516,20 +516,13 @@ func (r *Supervisor) Cancel(taskID string) {
 	<-delay.doneCh
 }
 
-// CancelAll aborts all pending restarts and waits for any instances of
-// StartNow that have already triggered to complete.
+// CancelAll aborts all pending restarts
 func (r *Supervisor) CancelAll() {
-	var cancelled []delayedStart
-
 	r.mu.Lock()
 	for _, delay := range r.delays {
 		delay.cancel()
 	}
 	r.mu.Unlock()
-
-	for _, delay := range cancelled {
-		<-delay.doneCh
-	}
 }
 
 // ClearServiceHistory forgets restart history related to a given service ID.

vendor/github.com/docker/swarmkit/manager/orchestrator/service.go

@@ -47,22 +47,27 @@ func SetServiceTasksRemove(ctx context.Context, s *store.MemoryStore, service *a
 	err = s.Batch(func(batch *store.Batch) error {
 		for _, t := range tasks {
 			err := batch.Update(func(tx store.Tx) error {
+				// the task may have changed for some reason in the meantime
+				// since we read it out, so we need to get from the store again
+				// within the boundaries of a transaction
+				latestTask := store.GetTask(tx, t.ID)
+
 				// time travel is not allowed. if the current desired state is
 				// above the one we're trying to go to we can't go backwards.
 				// we have nothing to do and we should skip to the next task
-				if t.DesiredState > api.TaskStateRemove {
+				if latestTask.DesiredState > api.TaskStateRemove {
 					// log a warning, though. we shouln't be trying to rewrite
 					// a state to an earlier state
 					log.G(ctx).Warnf(
 						"cannot update task %v in desired state %v to an earlier desired state %v",
-						t.ID, t.DesiredState, api.TaskStateRemove,
+						latestTask.ID, latestTask.DesiredState, api.TaskStateRemove,
 					)
 					return nil
 				}
 				// update desired state to REMOVE
-				t.DesiredState = api.TaskStateRemove
+				latestTask.DesiredState = api.TaskStateRemove
 
-				if err := store.UpdateTask(tx, t); err != nil {
+				if err := store.UpdateTask(tx, latestTask); err != nil {
 					log.G(ctx).WithError(err).Errorf("failed transaction: update task desired state to REMOVE")
 				}
 				return nil

vendor/github.com/docker/swarmkit/manager/orchestrator/update/updater.go

@@ -501,7 +501,10 @@ func (u *Updater) removeOldTasks(ctx context.Context, batch *store.Batch, remove
 				return fmt.Errorf("task %s not found while trying to shut it down", original.ID)
 			}
 			if t.DesiredState > api.TaskStateRunning {
-				return fmt.Errorf("task %s was already shut down when reached by updater", original.ID)
+				return fmt.Errorf(
+					"task %s was already shut down when reached by updater (state: %v)",
+					original.ID, t.DesiredState,
+				)
 			}
 			t.DesiredState = api.TaskStateShutdown
 			return store.UpdateTask(tx, t)

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

@@ -6,6 +6,7 @@ import (
 	"crypto/tls"
 	"encoding/json"
 	"io/ioutil"
+	"math"
 	"net"
 	"os"
 	"path/filepath"
@@ -33,7 +34,7 @@ import (
 	"github.com/docker/swarmkit/manager/encryption"
 	"github.com/docker/swarmkit/remotes"
 	"github.com/docker/swarmkit/xnet"
-	"github.com/grpc-ecosystem/go-grpc-prometheus"
+	grpc_prometheus "github.com/grpc-ecosystem/go-grpc-prometheus"
 	"github.com/pkg/errors"
 	"github.com/sirupsen/logrus"
 	bolt "go.etcd.io/bbolt"
@@ -911,6 +912,7 @@ func (n *Node) initManagerConnection(ctx context.Context, ready chan<- struct{})
 	opts := []grpc.DialOption{
 		grpc.WithUnaryInterceptor(grpc_prometheus.UnaryClientInterceptor),
 		grpc.WithStreamInterceptor(grpc_prometheus.StreamClientInterceptor),
+		grpc.WithDefaultCallOptions(grpc.MaxCallRecvMsgSize(math.MaxInt32)),
 	}
 	insecureCreds := credentials.NewTLS(&tls.Config{InsecureSkipVerify: true})
 	opts = append(opts, grpc.WithTransportCredentials(insecureCreds))