Workaround for inconsistency container status

Signed-off-by: Shawn Wang <[email protected]>

cmd/dockerd/daemon.go

@@ -274,10 +274,21 @@ func (cli *DaemonCli) start(opts daemonOptions) (err error) {
 	cli.TrustKeyPath = opts.common.TrustKey
 
 	registryService := registry.NewService(cli.Config.ServiceOptions)
-	containerdRemote, err := libcontainerd.New(cli.getLibcontainerdRoot(), cli.getPlatformRemoteOptions()...)
+
+	// libcontainerd.New add hook
+	hook := func() {
+		if cli.d == nil {
+			return
+		}
+		for _, c := range cli.d.List() {
+			c.CheckProcessIsRunning()
+		}
+	}
+	containerdRemote, err := libcontainerd.New(cli.getLibcontainerdRoot(), hook, cli.getPlatformRemoteOptions()...)
 	if err != nil {
 		return err
 	}
+
 	signal.Trap(func() {
 		cli.stop()
 		<-stopc // wait for daemonCli.start() to return

container/state.go

@@ -8,6 +8,7 @@ import (
 	"golang.org/x/net/context"
 
 	"github.com/docker/docker/api/types"
+	"github.com/docker/docker/utils"
 	"github.com/docker/go-units"
 )
 
@@ -230,6 +231,17 @@ func (s *State) IsRunning() bool {
 	return res
 }
 
+// CheckProcessIsRunning check the the process of running container is still performed.
+func (s *State) CheckProcessIsRunning() {
+	if !s.IsRunning() {
+		return
+	}
+	if !utils.IsProcessAlive(s.Pid) {
+		// TODO: force set exitcode:1, better to define new exitcode
+		s.SetStopped(&ExitStatus{ExitCode: 1})
+	}
+}
+
 // GetPID holds the process id of a container.
 func (s *State) GetPID() int {
 	s.Lock()

libcontainerd/remote_unix.go

@@ -61,10 +61,11 @@ type remote struct {
 	oomScore              int
 	maxHealthCheckRetries int
 	restoreFromTimestamp  *timestamp.Timestamp
+	handleContainerdHook  func()
 }
 
 // New creates a fresh instance of libcontainerd remote.
-func New(stateDir string, options ...RemoteOption) (_ Remote, err error) {
+func New(stateDir string, hook func(), options ...RemoteOption) (_ Remote, err error) {
 	defer func() {
 		if err != nil {
 			err = fmt.Errorf("Failed to connect to containerd. Please make sure containerd is installed in your PATH or you have specified the correct address. Got error: %v", err)
@@ -75,6 +76,7 @@ func New(stateDir string, options ...RemoteOption) (_ Remote, err error) {
 		daemonPid:   -1,
 		eventTsPath: filepath.Join(stateDir, eventTimestampFilename),
 	}
+	r.handleContainerdHook = hook
 	for _, option := range options {
 		if err := option.Apply(r); err != nil {
 			return nil, err
@@ -117,7 +119,6 @@ func New(stateDir string, options ...RemoteOption) (_ Remote, err error) {
 		logrus.Errorf("libcontainerd: failed to convert timestamp: %q", err)
 	}
 	r.restoreFromTimestamp = tsp
-
 	go r.handleConnectionChange()
 
 	if err := r.startEventsMonitor(); err != nil {
@@ -179,6 +180,10 @@ func (r *remote) handleConnectionChange() {
 				if err := r.runContainerdDaemon(); err != nil { //FIXME: Handle error
 					logrus.Errorf("libcontainerd: error restarting containerd: %v", err)
 				}
+				// Workaround to fix inconsistency stopped status
+				go func() {
+					r.handleContainerdHook()
+				}()
 				continue
 			}
 		}

libcontainerd/remote_windows.go

@@ -26,7 +26,7 @@ func (r *remote) UpdateOptions(opts ...RemoteOption) error {
 
 // New creates a fresh instance of libcontainerd remote. On Windows,
 // this is not used as there is no remote containerd process.
-func New(_ string, _ ...RemoteOption) (Remote, error) {
+func New(_ string, _ func(), _ ...RemoteOption) (Remote, error) {
 	return &remote{}, nil
 }