BACKPORT: Fix BZ#1529300

moby/moby#31662
moby/moby#31852

Signed-off-by: Antonio Murdaca <runcom@redhat.com>

containerd/main.go

@@ -20,7 +20,7 @@ import (
 	"github.com/codegangsta/cli"
 	"github.com/cyberdelia/go-metrics-graphite"
 	"github.com/docker/containerd"
-	"github.com/docker/containerd/api/grpc/server"
+	grpcserver "github.com/docker/containerd/api/grpc/server"
 	"github.com/docker/containerd/api/grpc/types"
 	"github.com/docker/containerd/api/http/pprof"
 	"github.com/docker/containerd/supervisor"
@@ -157,10 +157,25 @@ func main() {
 }
 
 func daemon(context *cli.Context) error {
+	stateDir := context.String("state-dir")
+	if err := os.MkdirAll(stateDir, 0755); err != nil {
+		return err
+	}
 	s := make(chan os.Signal, 2048)
 	signal.Notify(s, syscall.SIGTERM, syscall.SIGINT, syscall.SIGPIPE)
+	// Split the listen string of the form proto://addr
+	listenSpec := context.String("listen")
+	listenParts := strings.SplitN(listenSpec, "://", 2)
+	if len(listenParts) != 2 {
+		return fmt.Errorf("bad listen address format %s, expected proto://address", listenSpec)
+	}
+	// Register server early to allow healthcheck to be done
+	server, err := startServer(listenParts[0], listenParts[1])
+	if err != nil {
+		return err
+	}
 	sv, err := supervisor.New(
-		context.String("state-dir"),
+		stateDir,
 		context.String("runtime"),
 		context.String("shim"),
 		context.StringSlice("runtime-args"),
@@ -169,6 +184,7 @@ func daemon(context *cli.Context) error {
 	if err != nil {
 		return err
 	}
+	types.RegisterAPIServer(server, grpcserver.NewServer(sv))
 	wg := &sync.WaitGroup{}
 	for i := 0; i < 10; i++ {
 		wg.Add(1)
@@ -178,16 +194,6 @@ func daemon(context *cli.Context) error {
 	if err := sv.Start(); err != nil {
 		return err
 	}
-	// Split the listen string of the form proto://addr
-	listenSpec := context.String("listen")
-	listenParts := strings.SplitN(listenSpec, "://", 2)
-	if len(listenParts) != 2 {
-		return fmt.Errorf("bad listen address format %s, expected proto://address", listenSpec)
-	}
-	server, err := startServer(listenParts[0], listenParts[1], sv)
-	if err != nil {
-		return err
-	}
 	for ss := range s {
 		switch ss {
 		case syscall.SIGPIPE:
@@ -201,7 +207,7 @@ func daemon(context *cli.Context) error {
 	return nil
 }
 
-func startServer(protocol, address string, sv *supervisor.Supervisor) (*grpc.Server, error) {
+func startServer(protocol, address string) (*grpc.Server, error) {
 	// TODO: We should use TLS.
 	// TODO: Add an option for the SocketGroup.
 	sockets, err := listeners.Init(protocol, address, "", nil)
@@ -213,7 +219,6 @@ func startServer(protocol, address string, sv *supervisor.Supervisor) (*grpc.Ser
 	}
 	l := sockets[0]
 	s := grpc.NewServer()
-	types.RegisterAPIServer(s, server.NewServer(sv))
 	healthServer := health.NewHealthServer()
 	grpc_health_v1.RegisterHealthServer(s, healthServer)
 

supervisor/supervisor.go

@@ -21,9 +21,6 @@ const (
 // New returns an initialized Process supervisor.
 func New(stateDir string, runtimeName, shimName string, runtimeArgs []string, timeout time.Duration, retainCount int) (*Supervisor, error) {
 	startTasks := make(chan *startTask, 10)
-	if err := os.MkdirAll(stateDir, 0755); err != nil {
-		return nil, err
-	}
 	machine, err := CollectMachineInformation()
 	if err != nil {
 		return nil, err
@@ -334,7 +331,9 @@ func (s *Supervisor) restore() error {
 		id := d.Name()
 		container, err := runtime.Load(s.stateDir, id, s.shim, s.timeout)
 		if err != nil {
-			return err
+			logrus.WithFields(logrus.Fields{"error": err, "id": id}).Warnf("containerd: failed to load container,removing state directory.")
+			os.RemoveAll(filepath.Join(s.stateDir, id))
+			continue
 		}
 		processes, err := container.Processes()
 		if err != nil {