Merge pull request #1342 from mlaventure/is-serving-failfast

Disable FailFast on client.IsServing()

client.go

@@ -104,10 +104,14 @@ type Client struct {
 	runtime   string
 }
 
-// IsServing returns true if the client can successfully connect to the containerd daemon
-// and the healthcheck service returns the SERVING response
+// IsServing returns true if the client can successfully connect to the
+// containerd daemon and the healthcheck service returns the SERVING
+// response.
+// This call will block if a transient error is encountered during
+// connection. A timeout can be set in the context to ensure it returns
+// early.
 func (c *Client) IsServing(ctx context.Context) (bool, error) {
-	r, err := c.HealthService().Check(ctx, &grpc_health_v1.HealthCheckRequest{})
+	r, err := c.HealthService().Check(ctx, &grpc_health_v1.HealthCheckRequest{}, grpc.FailFast(false))
 	if err != nil {
 		return false, err
 	}

client_test.go

@@ -11,6 +11,7 @@ import (
 	"os/exec"
 	"runtime"
 	"testing"
+	"time"
 
 	"google.golang.org/grpc/grpclog"
 
@@ -73,18 +74,20 @@ func TestMain(m *testing.M) {
 		}
 	}
 
-	client, err := ctrd.waitForStart(ctx)
+	waitCtx, waitCancel := context.WithTimeout(ctx, 2*time.Second)
+	client, err := ctrd.waitForStart(waitCtx)
+	waitCancel()
 	if err != nil {
 		ctrd.Kill()
 		ctrd.Wait()
-		fmt.Fprintf(os.Stderr, "%s: %s", err, buf.String())
+		fmt.Fprintf(os.Stderr, "%s: %s\n", err, buf.String())
 		os.Exit(1)
 	}
 
 	// print out the version in information
 	version, err := client.Version(ctx)
 	if err != nil {
-		fmt.Fprintf(os.Stderr, "error getting version: %s", err)
+		fmt.Fprintf(os.Stderr, "error getting version: %s\n", err)
 		os.Exit(1)
 	}
 
@@ -99,7 +102,7 @@ func TestMain(m *testing.M) {
 		if _, err = client.Pull(ctx, testImage, WithPullUnpack); err != nil {
 			ctrd.Stop()
 			ctrd.Wait()
-			fmt.Fprintf(os.Stderr, "%s: %s", err, buf.String())
+			fmt.Fprintf(os.Stderr, "%s: %s\n", err, buf.String())
 			os.Exit(1)
 		}
 	}

client_windows.go

@@ -1,16 +0,0 @@
-package containerd
-
-import (
-	"net"
-	"time"
-
-	winio "github.com/Microsoft/go-winio"
-)
-
-func Dialer(address string, timeout time.Duration) (net.Conn, error) {
-	return winio.DialPipe(address, &timeout)
-}
-
-func DialAddress(address string) string {
-	return address
-}

container_linux_test.go

@@ -240,14 +240,9 @@ func TestDaemonRestart(t *testing.T) {
 
 	<-statusC
 
-	serving := false
-	for i := 0; i < 20; i++ {
-		serving, err = client.IsServing(ctx)
-		if serving {
-			break
-		}
-		time.Sleep(100 * time.Millisecond)
-	}
+	waitCtx, waitCancel := context.WithTimeout(ctx, 2*time.Second)
+	serving, err := client.IsServing(waitCtx)
+	waitCancel()
 	if !serving {
 		t.Fatalf("containerd did not start within 2s: %v", err)
 	}

daemon_test.go

@@ -2,7 +2,6 @@ package containerd
 
 import (
 	"context"
-	"fmt"
 	"io"
 	"os/exec"
 	"sync"
@@ -44,19 +43,19 @@ func (d *daemon) waitForStart(ctx context.Context) (*Client, error) {
 		err     error
 	)
 
-	for i := 0; i < 20; i++ {
-		if client == nil {
-			client, err = New(d.addr)
-		}
-		if err == nil {
-			serving, err = client.IsServing(ctx)
-			if serving {
-				return client, nil
-			}
-		}
-		time.Sleep(100 * time.Millisecond)
+	client, err = New(address)
+	if err != nil {
+		return nil, err
 	}
-	return nil, fmt.Errorf("containerd did not start within 2s: %v", err)
+	serving, err = client.IsServing(ctx)
+	if !serving {
+		client.Close()
+		if err == nil {
+			err = errors.New("connection was successful but service is not available")
+		}
+		return nil, err
+	}
+	return client, err
 }
 
 func (d *daemon) Stop() error {

dialer.go

@@ -1,31 +1,13 @@
-// +build !windows
-
 package containerd
 
 import (
-	"fmt"
 	"net"
-	"os"
 	"strings"
-	"syscall"
 	"time"
 
 	"github.com/pkg/errors"
 )
 
-func isNoent(err error) bool {
-	if err != nil {
-		if nerr, ok := err.(*net.OpError); ok {
-			if serr, ok := nerr.Err.(*os.SyscallError); ok {
-				if serr.Err == syscall.ENOENT {
-					return true
-				}
-			}
-		}
-	}
-	return false
-}
-
 type dialResult struct {
 	c   net.Conn
 	err error
@@ -44,7 +26,7 @@ func Dialer(address string, timeout time.Duration) (net.Conn, error) {
 			case <-stopC:
 				return
 			default:
-				c, err := net.DialTimeout("unix", address, timeout)
+				c, err := dialer(address, timeout)
 				if isNoent(err) {
 					<-time.After(10 * time.Millisecond)
 					continue
@@ -68,7 +50,3 @@ func Dialer(address string, timeout time.Duration) (net.Conn, error) {
 		return nil, errors.Errorf("dial %s: no such file or directory", address)
 	}
 }
-
-func DialAddress(address string) string {
-	return fmt.Sprintf("unix://%s", address)
-}

dialer_unix.go

@@ -0,0 +1,32 @@
+// +build !windows
+
+package containerd
+
+import (
+	"fmt"
+	"net"
+	"os"
+	"syscall"
+	"time"
+)
+
+func isNoent(err error) bool {
+	if err != nil {
+		if nerr, ok := err.(*net.OpError); ok {
+			if serr, ok := nerr.Err.(*os.SyscallError); ok {
+				if serr.Err == syscall.ENOENT {
+					return true
+				}
+			}
+		}
+	}
+	return false
+}
+
+func dialer(address string, timeout time.Duration) (net.Conn, error) {
+	return net.DialTimeout("unix", address, timeout)
+}
+
+func DialAddress(address string) string {
+	return fmt.Sprintf("unix://%s", address)
+}

dialer_windows.go

@@ -0,0 +1,29 @@
+package containerd
+
+import (
+	"net"
+	"os"
+	"syscall"
+	"time"
+
+	winio "github.com/Microsoft/go-winio"
+)
+
+func isNoent(err error) bool {
+	if err != nil {
+		if oerr, ok := err.(*os.PathError); ok {
+			if oerr.Err == syscall.ENOENT {
+				return true
+			}
+		}
+	}
+	return false
+}
+
+func dialer(address string, timeout time.Duration) (net.Conn, error) {
+	return winio.DialPipe(address, &timeout)
+}
+
+func DialAddress(address string) string {
+	return address
+}