Do not rely on "live" event anymore

Signed-off-by: Kenfe-Mickael Laventure <mickael.laventure@gmail.com>
(cherry picked from commit 64483c3bda)
Signed-off-by: Tibor Vass <tibor@docker.com>

daemon/daemon.go

@@ -176,7 +176,7 @@ func (daemon *Daemon) restore() error {
 			rm := c.RestartManager(false)
 			if c.IsRunning() || c.IsPaused() {
 				if err := daemon.containerd.Restore(c.ID, libcontainerd.WithRestartManager(rm)); err != nil {
-					logrus.Errorf("Failed to restore with containerd: %q", err)
+					logrus.Errorf("Failed to restore %s with containerd: %s", c.ID, err)
 					return
 				}
 				if !c.HostConfig.NetworkMode.IsContainer() && c.IsRunning() {

integration-cli/docker_cli_daemon_experimental_test.go

@@ -7,7 +7,7 @@ import (
 	"os/exec"
 	"path/filepath"
 	"strings"
-	"time"
+	"syscall"
 
 	"github.com/docker/docker/pkg/integration/checker"
 	"github.com/go-check/check"
@@ -129,7 +129,11 @@ func (s *DockerDaemonSuite) TestDaemonShutdownWithPlugins(c *check.C) {
 		c.Fatalf("Could not kill daemon: %v", err)
 	}
 
-	time.Sleep(5 * time.Second)
+	for {
+		if err := syscall.Kill(s.d.cmd.Process.Pid, 0); err == syscall.ESRCH {
+			break
+		}
+	}
 
 	cmd := exec.Command("pgrep", "-f", "plugin-no-remove")
 	if out, ec, err := runCommandWithOutput(cmd); ec != 1 {

libcontainerd/client_linux.go

@@ -281,16 +281,10 @@ func (clnt *client) cleanupOldRootfs(containerID string) {
 	}
 }
 
-func (clnt *client) setExited(containerID string) error {
+func (clnt *client) setExited(containerID string, exitCode uint32) error {
 	clnt.lock(containerID)
 	defer clnt.unlock(containerID)
 
-	var exitCode uint32
-	if event, ok := clnt.remote.pastEvents[containerID]; ok {
-		exitCode = event.Status
-		delete(clnt.remote.pastEvents, containerID)
-	}
-
 	err := clnt.backend.StateChanged(containerID, StateInfo{
 		CommonStateInfo: CommonStateInfo{
 			State:    StateExit,
@@ -393,7 +387,7 @@ func (clnt *client) getOrCreateExitNotifier(containerID string) *exitNotifier {
 	return w
 }
 
-func (clnt *client) restore(cont *containerd.Container, options ...CreateOption) (err error) {
+func (clnt *client) restore(cont *containerd.Container, lastEvent *containerd.Event, options ...CreateOption) (err error) {
 	clnt.lock(cont.Id)
 	defer clnt.unlock(cont.Id)
 
@@ -441,36 +435,103 @@ func (clnt *client) restore(cont *containerd.Container, options ...CreateOption)
 		return err
 	}
 
-	if event, ok := clnt.remote.pastEvents[containerID]; ok {
+	if lastEvent != nil {
 		// This should only be a pause or resume event
-		if event.Type == StatePause || event.Type == StateResume {
+		if lastEvent.Type == StatePause || lastEvent.Type == StateResume {
 			return clnt.backend.StateChanged(containerID, StateInfo{
 				CommonStateInfo: CommonStateInfo{
-					State: event.Type,
+					State: lastEvent.Type,
 					Pid:   container.systemPid,
 				}})
 		}
 
-		logrus.Warnf("unexpected backlog event: %#v", event)
+		logrus.Warnf("unexpected backlog event: %#v", lastEvent)
 	}
 
 	return nil
 }
 
+func (clnt *client) getContainerLastEvent(containerID string) (*containerd.Event, error) {
+	er := &containerd.EventsRequest{
+		Timestamp:  clnt.remote.restoreFromTimestamp,
+		StoredOnly: true,
+		Id:         containerID,
+	}
+	events, err := clnt.remote.apiClient.Events(context.Background(), er)
+	if err != nil {
+		logrus.Errorf("libcontainerd: failed to get container events stream for %s: %q", er.Id, err)
+		return nil, err
+	}
+
+	var ev *containerd.Event
+	for {
+		e, err := events.Recv()
+		if err != nil {
+			if err.Error() == "EOF" {
+				break
+			}
+			logrus.Errorf("libcontainerd: failed to get container event for %s: %q", containerID, err)
+			return nil, err
+		}
+
+		logrus.Debugf("libcontainerd: received past event %#v", e)
+
+		switch e.Type {
+		case StateExit, StatePause, StateResume:
+			ev = e
+		}
+	}
+
+	return ev, nil
+}
+
 func (clnt *client) Restore(containerID string, options ...CreateOption) error {
-	if clnt.liveRestore {
+	// Synchronize with live events
+	clnt.remote.Lock()
+	defer clnt.remote.Unlock()
+	// Check that containerd still knows this container.
+	//
+	// In the unlikely event that Restore for this container process
+	// the its past event before the main loop, the event will be
+	// processed twice. However, this is not an issue as all those
+	// events will do is change the state of the container to be
+	// exactly the same.
 	cont, err := clnt.getContainerdContainer(containerID)
-		if err == nil && cont.Status != "stopped" {
-			if err := clnt.restore(cont, options...); err != nil {
+	// Get its last event
+	ev, eerr := clnt.getContainerLastEvent(containerID)
+	if err != nil || cont.Status == "Stopped" {
+		if err != nil && !strings.Contains(err.Error(), "container not found") {
+			// Legitimate error
+			return err
+		}
+
+		// If ev is nil, then we already consumed all the event of the
+		// container, included the "exit" one.
+		// Thus we return to avoid overriding the Exit Code.
+		if ev == nil {
+			logrus.Warnf("libcontainerd: restore was called on a fully synced container (%s)", containerID)
+			return nil
+		}
+
+		// get the exit status for this container
+		ec := uint32(0)
+		if eerr == nil && ev.Type == StateExit {
+			ec = ev.Status
+		}
+		clnt.setExited(containerID, ec)
+
+		return nil
+	}
+
+	// container is still alive
+	if clnt.liveRestore {
+		if err := clnt.restore(cont, ev, options...); err != nil {
 			logrus.Errorf("error restoring %s: %v", containerID, err)
 		}
 		return nil
 	}
-		return clnt.setExited(containerID)
-	}
 
-	cont, err := clnt.getContainerdContainer(containerID)
-	if err == nil && cont.Status != "stopped" {
+	// Kill the container if liveRestore == false
 	w := clnt.getOrCreateExitNotifier(containerID)
 	clnt.lock(cont.Id)
 	container := clnt.newContainer(cont.BundlePath)
@@ -496,11 +557,10 @@ func (clnt *client) Restore(containerID string, options ...CreateOption) error {
 	case <-w.wait():
 		return nil
 	}
-	}
 
 	clnt.deleteContainer(containerID)
 
-	return clnt.setExited(containerID)
+	return clnt.setExited(containerID, uint32(255))
 }
 
 type exitNotifier struct {

libcontainerd/remote_linux.go

@@ -21,6 +21,7 @@ import (
 	sysinfo "github.com/docker/docker/pkg/system"
 	"github.com/docker/docker/utils"
 	"github.com/golang/protobuf/ptypes"
+	"github.com/golang/protobuf/ptypes/timestamp"
 	"golang.org/x/net/context"
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/grpclog"
@@ -50,12 +51,12 @@ type remote struct {
 	rpcConn              *grpc.ClientConn
 	clients              []*client
 	eventTsPath          string
-	pastEvents    map[string]*containerd.Event
 	runtime              string
 	runtimeArgs          []string
 	daemonWaitCh         chan struct{}
 	liveRestore          bool
 	oomScore             int
+	restoreFromTimestamp *timestamp.Timestamp
 }
 
 // New creates a fresh instance of libcontainerd remote.
@@ -69,7 +70,6 @@ func New(stateDir string, options ...RemoteOption) (_ Remote, err error) {
 		stateDir:    stateDir,
 		daemonPid:   -1,
 		eventTsPath: filepath.Join(stateDir, eventTimestampFilename),
-		pastEvents:  make(map[string]*containerd.Event),
 	}
 	for _, option := range options {
 		if err := option.Apply(r); err != nil {
@@ -106,6 +106,14 @@ func New(stateDir string, options ...RemoteOption) (_ Remote, err error) {
 	r.rpcConn = conn
 	r.apiClient = containerd.NewAPIClient(conn)
 
+	// Get the timestamp to restore from
+	t := r.getLastEventTimestamp()
+	tsp, err := ptypes.TimestampProto(t)
+	if err != nil {
+		logrus.Errorf("libcontainerd: failed to convert timestamp: %q", err)
+	}
+	r.restoreFromTimestamp = tsp
+
 	go r.handleConnectionChange()
 
 	if err := r.startEventsMonitor(); err != nil {
@@ -257,7 +265,8 @@ func (r *remote) getLastEventTimestamp() time.Time {
 
 func (r *remote) startEventsMonitor() error {
 	// First, get past events
-	tsp, err := ptypes.TimestampProto(r.getLastEventTimestamp())
+	t := r.getLastEventTimestamp()
+	tsp, err := ptypes.TimestampProto(t)
 	if err != nil {
 		logrus.Errorf("libcontainerd: failed to convert timestamp: %q", err)
 	}
@@ -299,7 +308,7 @@ func (r *remote) handleEventStream(events containerd.API_EventsClient) {
 		}
 		r.RUnlock()
 		if container == nil {
-			logrus.Errorf("libcontainerd: %q", err)
+			logrus.Warnf("libcontainerd: unknown container %s", e.Id)
 			continue
 		}