auto-update: simple rollback

Add support for simple rollbacks during `podman auto-update`.  Rollbacks
are enabled by default.  If a systemd unit cannot be restarted after an
update, the previous image will be retagged and the unit will be
restarted a second time.

Add system tests for rollbacks.  Also fix a bug in the restart sequence;
we have to use the channel to actually know whether the restart was
successful or not.

NOTE: To make rollbacks really useful, users must run their containers
with `--sdnotify=container` such that the containers send the ready
message over the (mounted) socket.  This way, restarting the systemd
units during auto update will block until the message has been received
(or a timeout kicked in).

Signed-off-by: Valentin Rothberg <rothberg@redhat.com>

5 files changed, 126 insertions, 38 deletions

diff --git a/pkg/autoupdate/autoupdate.go b/pkg/autoupdate/autoupdate.go
index c51e2cd03..b1ebbfa8e 100644
--- a/pkg/autoupdate/autoupdate.go
+++ b/pkg/autoupdate/autoupdate.go
@@ -88,7 +88,7 @@ func ValidateImageReference(imageName string) error {
 	} else if err != nil {
 		repo, err := reference.Parse(imageName)
 		if err != nil {
-			return errors.Wrap(err, "error enforcing fully-qualified docker transport reference for auto updates")
+			return errors.Wrap(err, "enforcing fully-qualified docker transport reference for auto updates")
 		}
 		if _, ok := repo.(reference.NamedTagged); !ok {
 			return errors.Errorf("auto updates require fully-qualified image references (no tag): %q", imageName)
@@ -181,13 +181,13 @@ func autoUpdateRegistry(ctx context.Context, image *libimage.Image, ctr *libpod.
 	cid := ctr.ID()
 	rawImageName := ctr.RawImageName()
 	if rawImageName == "" {
-		return nil, errors.Errorf("error registry auto-updating container %q: raw-image name is empty", cid)
+		return nil, errors.Errorf("registry auto-updating container %q: raw-image name is empty", cid)
 	}
 
 	labels := ctr.Labels()
 	unit, exists := labels[systemdDefine.EnvVariable]
 	if !exists {
-		return nil, errors.Errorf("error auto-updating container %q: no %s label found", ctr.ID(), systemdDefine.EnvVariable)
+		return nil, errors.Errorf("auto-updating container %q: no %s label found", ctr.ID(), systemdDefine.EnvVariable)
 	}
 
 	report := &entities.AutoUpdateReport{
@@ -201,7 +201,7 @@ func autoUpdateRegistry(ctx context.Context, image *libimage.Image, ctr *libpod.
 
 	if _, updated := updatedRawImages[rawImageName]; updated {
 		logrus.Infof("Auto-updating container %q using registry image %q", cid, rawImageName)
-		if err := restartSystemdUnit(ctr, unit, conn); err != nil {
+		if err := restartSystemdUnit(ctx, ctr, unit, conn); err != nil {
 			return report, err
 		}
 		report.Updated = "true"
@@ -211,7 +211,7 @@ func autoUpdateRegistry(ctx context.Context, image *libimage.Image, ctr *libpod.
 	authfile := getAuthfilePath(ctr, options)
 	needsUpdate, err := newerRemoteImageAvailable(ctx, runtime, image, rawImageName, authfile)
 	if err != nil {
-		return report, errors.Wrapf(err, "error registry auto-updating container %q: image check for %q failed", cid, rawImageName)
+		return report, errors.Wrapf(err, "registry auto-updating container %q: image check for %q failed", cid, rawImageName)
 	}
 
 	if !needsUpdate {
@@ -225,16 +225,30 @@ func autoUpdateRegistry(ctx context.Context, image *libimage.Image, ctr *libpod.
 	}
 
 	if _, err := updateImage(ctx, runtime, rawImageName, options); err != nil {
-		return report, errors.Wrapf(err, "error registry auto-updating container %q: image update for %q failed", cid, rawImageName)
+		return report, errors.Wrapf(err, "registry auto-updating container %q: image update for %q failed", cid, rawImageName)
 	}
 	updatedRawImages[rawImageName] = true
 
 	logrus.Infof("Auto-updating container %q using registry image %q", cid, rawImageName)
-	if err := restartSystemdUnit(ctr, unit, conn); err != nil {
-		return report, err
+	updateErr := restartSystemdUnit(ctx, ctr, unit, conn)
+	if updateErr == nil {
+		report.Updated = "true"
+		return report, nil
+	}
+
+	if !options.Rollback {
+		return report, updateErr
+	}
+
+	// To fallback, simply retag the old image and restart the service.
+	if err := image.Tag(rawImageName); err != nil {
+		return report, errors.Wrap(err, "falling back to previous image")
+	}
+	if err := restartSystemdUnit(ctx, ctr, unit, conn); err != nil {
+		return report, errors.Wrap(err, "restarting unit with old image during fallback")
 	}
 
-	report.Updated = "true"
+	report.Updated = "rolled back"
 	return report, nil
 }
 
@@ -243,13 +257,13 @@ func autoUpdateLocally(ctx context.Context, image *libimage.Image, ctr *libpod.C
 	cid := ctr.ID()
 	rawImageName := ctr.RawImageName()
 	if rawImageName == "" {
-		return nil, errors.Errorf("error locally auto-updating container %q: raw-image name is empty", cid)
+		return nil, errors.Errorf("locally auto-updating container %q: raw-image name is empty", cid)
 	}
 
 	labels := ctr.Labels()
 	unit, exists := labels[systemdDefine.EnvVariable]
 	if !exists {
-		return nil, errors.Errorf("error auto-updating container %q: no %s label found", ctr.ID(), systemdDefine.EnvVariable)
+		return nil, errors.Errorf("auto-updating container %q: no %s label found", ctr.ID(), systemdDefine.EnvVariable)
 	}
 
 	report := &entities.AutoUpdateReport{
@@ -263,7 +277,7 @@ func autoUpdateLocally(ctx context.Context, image *libimage.Image, ctr *libpod.C
 
 	needsUpdate, err := newerLocalImageAvailable(runtime, image, rawImageName)
 	if err != nil {
-		return report, errors.Wrapf(err, "error locally auto-updating container %q: image check for %q failed", cid, rawImageName)
+		return report, errors.Wrapf(err, "locally auto-updating container %q: image check for %q failed", cid, rawImageName)
 	}
 
 	if !needsUpdate {
@@ -277,23 +291,47 @@ func autoUpdateLocally(ctx context.Context, image *libimage.Image, ctr *libpod.C
 	}
 
 	logrus.Infof("Auto-updating container %q using local image %q", cid, rawImageName)
-	if err := restartSystemdUnit(ctr, unit, conn); err != nil {
-		return report, err
+	updateErr := restartSystemdUnit(ctx, ctr, unit, conn)
+	if updateErr == nil {
+		report.Updated = "true"
+		return report, nil
 	}
 
-	report.Updated = "true"
+	if !options.Rollback {
+		return report, updateErr
+	}
+
+	// To fallback, simply retag the old image and restart the service.
+	if err := image.Tag(rawImageName); err != nil {
+		return report, errors.Wrap(err, "falling back to previous image")
+	}
+	if err := restartSystemdUnit(ctx, ctr, unit, conn); err != nil {
+		return report, errors.Wrap(err, "restarting unit with old image during fallback")
+	}
+
+	report.Updated = "rolled back"
 	return report, nil
 }
 
 // restartSystemdUnit restarts the systemd unit the container is running in.
-func restartSystemdUnit(ctr *libpod.Container, unit string, conn *dbus.Conn) error {
-	_, err := conn.RestartUnit(unit, "replace", nil)
-	if err != nil {
-		return errors.Wrapf(err, "error auto-updating container %q: restarting systemd unit %q failed", ctr.ID(), unit)
+func restartSystemdUnit(ctx context.Context, ctr *libpod.Container, unit string, conn *dbus.Conn) error {
+	restartChan := make(chan string)
+	if _, err := conn.RestartUnitContext(ctx, unit, "replace", restartChan); err != nil {
+		return errors.Wrapf(err, "auto-updating container %q: restarting systemd unit %q failed", ctr.ID(), unit)
 	}
 
-	logrus.Infof("Successfully restarted systemd unit %q of container %q", unit, ctr.ID())
-	return nil
+	// Wait for the restart to finish and actually check if it was
+	// successful or not.
+	result := <-restartChan
+
+	switch result {
+	case "done":
+		logrus.Infof("Successfully restarted systemd unit %q of container %q", unit, ctr.ID())
+		return nil
+
+	default:
+		return errors.Errorf("auto-updating container %q: restarting systemd unit %q failed: expected %q but received %q", ctr.ID(), unit, "done", result)
+	}
 }
 
 // imageContainersMap generates a map[image ID] -> [containers using the image]
diff --git a/pkg/domain/entities/auto-update.go b/pkg/domain/entities/auto-update.go
index eed617bf8..5ea2cdf15 100644
--- a/pkg/domain/entities/auto-update.go
+++ b/pkg/domain/entities/auto-update.go
@@ -8,6 +8,9 @@ type AutoUpdateOptions struct {
 	// pending, it will be indicated in the Updated field of
 	// AutoUpdateReport.
 	DryRun bool
+	// If restarting the service with the new image failed, restart it
+	// another time with the previous image.
+	Rollback bool
 }
 
 // AutoUpdateReport contains the results from running auto-update.
diff --git a/pkg/systemd/generate/common.go b/pkg/systemd/generate/common.go
index 349805980..45e12014a 100644
--- a/pkg/systemd/generate/common.go
+++ b/pkg/systemd/generate/common.go
@@ -74,8 +74,7 @@ func filterCommonContainerFlags(command []string, argCount int) []string {
 		case s == "--sdnotify", s == "--cgroups":
 			i++
 			continue
-		case strings.HasPrefix(s, "--sdnotify="),
-			strings.HasPrefix(s, "--rm="),
+		case strings.HasPrefix(s, "--rm="),
 			strings.HasPrefix(s, "--cgroups="):
 			continue
 		}
diff --git a/pkg/systemd/generate/containers.go b/pkg/systemd/generate/containers.go
index 083520316..78b81b54b 100644
--- a/pkg/systemd/generate/containers.go
+++ b/pkg/systemd/generate/containers.go
@@ -258,7 +258,6 @@ func executeContainerTemplate(info *containerInfo, options entities.GenerateSyst
 		}
 		startCommand = append(startCommand,
 			"run",
-			"--sdnotify=conmon",
 			"--cgroups=no-conmon",
 			"--rm",
 		)
@@ -273,6 +272,7 @@ func executeContainerTemplate(info *containerInfo, options entities.GenerateSyst
 		fs.String("name", "", "")
 		fs.Bool("replace", false, "")
 		fs.StringArrayP("env", "e", nil, "")
+		fs.String("sdnotify", "", "")
 		fs.Parse(remainingCmd)
 
 		remainingCmd = filterCommonContainerFlags(remainingCmd, fs.NArg())
@@ -294,6 +294,13 @@ func executeContainerTemplate(info *containerInfo, options entities.GenerateSyst
 			return "", err
 		}
 
+		// Default to --sdnotify=conmon unless already set by the
+		// container.
+		hasSdnotifyParam := fs.Lookup("sdnotify").Changed
+		if !hasSdnotifyParam {
+			startCommand = append(startCommand, "--sdnotify=conmon")
+		}
+
 		if !hasDetachParam {
 			// Enforce detaching
 			//
diff --git a/pkg/systemd/generate/containers_test.go b/pkg/systemd/generate/containers_test.go
index 1d24cc4a9..6141950d0 100644
--- a/pkg/systemd/generate/containers_test.go
+++ b/pkg/systemd/generate/containers_test.go
@@ -130,7 +130,29 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman container run --sdnotify=conmon --cgroups=no-conmon --rm -d --replace --name jadda-jadda --hostname hello-world awesome-image:latest command arg1 ... argN "foo=arg \"with \" space"
+ExecStart=/usr/bin/podman container run --cgroups=no-conmon --rm --sdnotify=conmon -d --replace --name jadda-jadda --hostname hello-world awesome-image:latest command arg1 ... argN "foo=arg \"with \" space"
+Type=notify
+NotifyAccess=all
+
+[Install]
+WantedBy=multi-user.target default.target
+`
+
+	goodWithNameAndSdnotify := `# jadda-jadda.service
+# autogenerated by Podman CI
+
+[Unit]
+Description=Podman jadda-jadda.service
+Documentation=man:podman-generate-systemd(1)
+Wants=network-online.target
+After=network-online.target
+RequiresMountsFor=/var/run/containers/storage
+
+[Service]
+Environment=PODMAN_SYSTEMD_UNIT=%n
+Restart=always
+TimeoutStopSec=70
+ExecStart=/usr/bin/podman container run --cgroups=no-conmon --rm -d --replace --sdnotify=container --name jadda-jadda --hostname hello-world awesome-image:latest command arg1 ... argN "foo=arg \"with \" space"
 Type=notify
 NotifyAccess=all
 
@@ -152,7 +174,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm --replace -d --name jadda-jadda --hostname hello-world awesome-image:latest command arg1 ... argN
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --sdnotify=conmon --replace -d --name jadda-jadda --hostname hello-world awesome-image:latest command arg1 ... argN
 Type=notify
 NotifyAccess=all
 
@@ -174,7 +196,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm --pod-id-file %t/pod-foobar.pod-id-file --replace -d --name jadda-jadda --hostname hello-world awesome-image:latest command arg1 ... argN
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --pod-id-file %t/pod-foobar.pod-id-file --sdnotify=conmon --replace -d --name jadda-jadda --hostname hello-world awesome-image:latest command arg1 ... argN
 Type=notify
 NotifyAccess=all
 
@@ -196,7 +218,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm --replace --detach --name jadda-jadda --hostname hello-world awesome-image:latest command arg1 ... argN
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --sdnotify=conmon --replace --detach --name jadda-jadda --hostname hello-world awesome-image:latest command arg1 ... argN
 Type=notify
 NotifyAccess=all
 
@@ -218,7 +240,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm -d awesome-image:latest
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --sdnotify=conmon -d awesome-image:latest
 Type=notify
 NotifyAccess=all
 
@@ -241,7 +263,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=102
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm ` +
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --sdnotify=conmon ` +
 			detachparam +
 			` awesome-image:latest
 Type=notify
@@ -267,7 +289,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=102
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm -d --replace --name test -p 80:80 awesome-image:latest somecmd --detach=false
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --sdnotify=conmon -d --replace --name test -p 80:80 awesome-image:latest somecmd --detach=false
 Type=notify
 NotifyAccess=all
 
@@ -289,7 +311,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=102
-ExecStart=/usr/bin/podman --events-backend none --runroot /root run --sdnotify=conmon --cgroups=no-conmon --rm -d awesome-image:latest
+ExecStart=/usr/bin/podman --events-backend none --runroot /root run --cgroups=no-conmon --rm --sdnotify=conmon -d awesome-image:latest
 Type=notify
 NotifyAccess=all
 
@@ -311,7 +333,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman container run --sdnotify=conmon --cgroups=no-conmon --rm -d awesome-image:latest
+ExecStart=/usr/bin/podman container run --cgroups=no-conmon --rm --sdnotify=conmon -d awesome-image:latest
 Type=notify
 NotifyAccess=all
 
@@ -333,7 +355,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm -d --replace --name test --log-driver=journald --log-opt=tag={{.Name}} awesome-image:latest
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --sdnotify=conmon -d --replace --name test --log-driver=journald --log-opt=tag={{.Name}} awesome-image:latest
 Type=notify
 NotifyAccess=all
 
@@ -355,7 +377,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm -d --replace --name test awesome-image:latest sh -c "kill $$$$ && echo %%\\"
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --sdnotify=conmon -d --replace --name test awesome-image:latest sh -c "kill $$$$ && echo %%\\"
 Type=notify
 NotifyAccess=all
 
@@ -377,7 +399,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm -d --conmon-pidfile=foo --cidfile=foo awesome-image:latest podman run --cgroups=foo --conmon-pidfile=foo --cidfile=foo alpine
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --sdnotify=conmon -d --conmon-pidfile=foo --cidfile=foo awesome-image:latest podman run --cgroups=foo --conmon-pidfile=foo --cidfile=foo alpine
 Type=notify
 NotifyAccess=all
 
@@ -399,7 +421,7 @@ RequiresMountsFor=/var/run/containers/storage
 Environment=PODMAN_SYSTEMD_UNIT=%n
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm --pod-id-file %t/pod-foobar.pod-id-file -d --conmon-pidfile=foo --cidfile=foo awesome-image:latest podman run --cgroups=foo --conmon-pidfile=foo --cidfile=foo --pod-id-file /tmp/pod-foobar.pod-id-file alpine
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --pod-id-file %t/pod-foobar.pod-id-file --sdnotify=conmon -d --conmon-pidfile=foo --cidfile=foo awesome-image:latest podman run --cgroups=foo --conmon-pidfile=foo --cidfile=foo --pod-id-file /tmp/pod-foobar.pod-id-file alpine
 Type=notify
 NotifyAccess=all
 
@@ -422,7 +444,7 @@ Environment=PODMAN_SYSTEMD_UNIT=%n
 Environment=FOO=abc "BAR=my test" USER=%%a
 Restart=always
 TimeoutStopSec=70
-ExecStart=/usr/bin/podman run --sdnotify=conmon --cgroups=no-conmon --rm -d --env FOO --env=BAR --env=MYENV=2 -e USER awesome-image:latest
+ExecStart=/usr/bin/podman run --cgroups=no-conmon --rm --sdnotify=conmon -d --env FOO --env=BAR --env=MYENV=2 -e USER awesome-image:latest
 Type=notify
 NotifyAccess=all
 
@@ -547,6 +569,25 @@ WantedBy=multi-user.target default.target
 			false,
 			false,
 		},
+		{"good with name and sdnotify",
+			containerInfo{
+				Executable:        "/usr/bin/podman",
+				ServiceName:       "jadda-jadda",
+				ContainerNameOrID: "jadda-jadda",
+				RestartPolicy:     "always",
+				PIDFile:           "/run/containers/storage/overlay-containers/639c53578af4d84b8800b4635fa4e680ee80fd67e0e6a2d4eea48d1e3230f401/userdata/conmon.pid",
+				StopTimeout:       10,
+				PodmanVersion:     "CI",
+				CreateCommand:     []string{"I'll get stripped", "container", "run", "--sdnotify=container", "--name", "jadda-jadda", "--hostname", "hello-world", "awesome-image:latest", "command", "arg1", "...", "argN", "foo=arg \"with \" space"},
+				EnvVariable:       define.EnvVariable,
+				GraphRoot:         "/var/lib/containers/storage",
+				RunRoot:           "/var/run/containers/storage",
+			},
+			goodWithNameAndSdnotify,
+			true,
+			false,
+			false,
+		},
 		{"good with explicit short detach param",
 			containerInfo{
 				Executable:        "/usr/bin/podman",