Implement container restarting

Signed-off-by: Matthew Heon <matthew.heon@gmail.com>

Closes: #482
Approved by: baude

5 files changed, 101 insertions, 28 deletions

diff --git a/cmd/podman/start.go b/cmd/podman/start.go
index 18e8f7766..243fe71e2 100644
--- a/cmd/podman/start.go
+++ b/cmd/podman/start.go
@@ -104,8 +104,18 @@ func startCmd(c *cli.Context) error {
 		if err != nil {
 			return errors.Wrapf(err, "unable to parse annotations in %s", ctr.ID())
 		}
+		err = ctr.Start()
+		if err != nil {
+			if lastError != nil {
+				fmt.Fprintln(os.Stderr, lastError)
+			}
+			lastError = errors.Wrapf(err, "unable to start %s", container)
+			continue
+		}
 		// We only get a terminal session if both a tty was specified in the spec and
 		// -a on the command-line was given.
+		// Must be done after Start() because we might be restarting
+		// If so, the attach socket might be removed & recreated
 		if attach && tty {
 			// We increment the wg counter because we need to do the attach
 			wg.Add(1)
@@ -121,14 +131,6 @@ func startCmd(c *cli.Context) error {
 				return errors.Errorf("unable to attach to container %s", ctr.ID())
 			}
 		}
-		err = ctr.Start()
-		if err != nil {
-			if lastError != nil {
-				fmt.Fprintln(os.Stderr, lastError)
-			}
-			lastError = errors.Wrapf(err, "unable to start %s", container)
-			continue
-		}
 		if !attach {
 			fmt.Println(ctr.ID())
 		}
@@ -141,7 +143,11 @@ func startCmd(c *cli.Context) error {
 		if lastError != nil {
 			fmt.Fprintln(os.Stderr, lastError)
 		}
-		lastError = ctr.Cleanup()
+		// We can only do this if we attached
+		// Otherwise the container is probably still running
+		if attach && tty {
+			lastError = ctr.Cleanup()
+		}
 	}
 	return lastError
 }
diff --git a/libpod/container.go b/libpod/container.go
index 648dc821e..98810ea94 100644
--- a/libpod/container.go
+++ b/libpod/container.go
@@ -106,8 +106,6 @@ type Container struct {
 	runtime *Runtime
 }
 
-// TODO fetch IP and Subnet Mask from networks once we have updated OCICNI
-
 // containerState contains the current state of the container
 // It is stored on disk in a tmpfs and recreated on reboot
 type containerState struct {
diff --git a/libpod/container_api.go b/libpod/container_api.go
index 7e2197614..499cf826f 100644
--- a/libpod/container_api.go
+++ b/libpod/container_api.go
@@ -63,6 +63,9 @@ func (c *Container) Init() (err error) {
 }
 
 // Start starts a container
+// Start can start created or stopped containers
+// Stopped containers will be deleted and re-created in runc, undergoing a fresh
+// Init()
 func (c *Container) Start() (err error) {
 	if !c.locked {
 		c.lock.Lock()
@@ -78,11 +81,6 @@ func (c *Container) Start() (err error) {
 		return errors.Wrapf(ErrCtrStateInvalid, "container %s must be in Created or Stopped state to be started", c.ID())
 	}
 
-	// TODO remove this when we patch conmon to support restarting containers
-	if c.state.State == ContainerStateStopped {
-		return errors.Wrapf(ErrNotImplemented, "restarting a stopped container is not yet supported")
-	}
-
 	// Mount storage for the container if necessary
 	if err := c.mountStorage(); err != nil {
 		return err
@@ -109,6 +107,34 @@ func (c *Container) Start() (err error) {
 		}
 	}()
 
+	// Reinitialize the container if we need to
+	if c.state.State == ContainerStateStopped {
+		// If necessary, delete attach and ctl files
+		if err := c.removeConmonFiles(); err != nil {
+			return err
+		}
+
+		// Delete the container in the runtime
+		if err := c.runtime.ociRuntime.deleteContainer(c); err != nil {
+			return errors.Wrapf(err, "error removing container %s from runtime", c.ID())
+		}
+
+		// Our state is now Configured, as we've removed ourself from
+		// the runtime
+		// Set and save now to make sure that, if the init() below fails
+		// we still have a valid state
+		c.state.State = ContainerStateConfigured
+		if err := c.save(); err != nil {
+			return err
+		}
+
+		// Reinitialize the container
+		if err := c.init(); err != nil {
+			return err
+		}
+	}
+
+	// Start the container
 	return c.start()
 }
 
diff --git a/libpod/container_internal.go b/libpod/container_internal.go
index 85c4283b3..6b375ef6f 100644
--- a/libpod/container_internal.go
+++ b/libpod/container_internal.go
@@ -255,6 +255,38 @@ func (c *Container) refresh() error {
 		return errors.Wrapf(err, "error refreshing state for container %s", c.ID())
 	}
 
+	// Remove ctl and attach files, which may persist across reboot
+	if err := c.removeConmonFiles(); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+// Remove conmon attach socket and terminal resize FIFO
+// This is necessary for restarting containers
+func (c *Container) removeConmonFiles() error {
+	// Files are allowed to not exist, so ignore ENOENT
+	attachFile := filepath.Join(c.bundlePath(), "attach")
+	if err := os.Remove(attachFile); err != nil && !os.IsNotExist(err) {
+		return errors.Wrapf(err, "error removing container %s attach file", c.ID())
+	}
+
+	ctlFile := filepath.Join(c.bundlePath(), "ctl")
+	if err := os.Remove(ctlFile); err != nil && !os.IsNotExist(err) {
+		return errors.Wrapf(err, "error removing container %s ctl file", c.ID())
+	}
+
+	oomFile := filepath.Join(c.bundlePath(), "oom")
+	if err := os.Remove(oomFile); err != nil && !os.IsNotExist(err) {
+		return errors.Wrapf(err, "error removing container %s OOM file", c.ID())
+	}
+
+	exitFile := filepath.Join(c.runtime.ociRuntime.exitsDir, c.ID())
+	if err := os.Remove(exitFile); err != nil && !os.IsNotExist(err) {
+		return errors.Wrapf(err, "error removing container %s exit file", c.ID())
+	}
+
 	return nil
 }
 
@@ -360,10 +392,6 @@ func (c *Container) initAndStart() (err error) {
 	if c.state.State == ContainerStatePaused {
 		return errors.Wrapf(ErrCtrStateInvalid, "cannot start paused container %s", c.ID())
 	}
-	// TODO remove this once we can restart containers
-	if c.state.State == ContainerStateStopped {
-		return errors.Wrapf(ErrNotImplemented, "restarting containers is not yet implemented")
-	}
 
 	// Mount if necessary
 	if err := c.mountStorage(); err != nil {
@@ -391,6 +419,29 @@ func (c *Container) initAndStart() (err error) {
 		}
 	}()
 
+	// If we are ContainerStateStopped we need to remove from runtime
+	// And reset to ContainerStateConfigured
+	if c.state.State == ContainerStateStopped {
+		// If necessary, delete attach and ctl files
+		if err := c.removeConmonFiles(); err != nil {
+			return err
+		}
+
+		// Delete the container in the runtime
+		if err := c.runtime.ociRuntime.deleteContainer(c); err != nil {
+			return errors.Wrapf(err, "error removing container %s from runtime", c.ID())
+		}
+
+		// Our state is now Configured, as we've removed ourself from
+		// the runtime
+		// Set and save now to make sure that, if the init() below fails
+		// we still have a valid state
+		c.state.State = ContainerStateConfigured
+		if err := c.save(); err != nil {
+			return err
+		}
+	}
+
 	// If we are ContainerStateConfigured we need to init()
 	if c.state.State == ContainerStateConfigured {
 		if err := c.init(); err != nil {
diff --git a/libpod/pod.go b/libpod/pod.go
index 560277578..bd02efdae 100644
--- a/libpod/pod.go
+++ b/libpod/pod.go
@@ -130,14 +130,6 @@ func (p *Pod) Start() (map[string]error, error) {
 	for len(ctrsToStart) > 0 {
 		// Loop through all containers, attempting to start them
 		for id, ctr := range ctrsToStart {
-			// TODO remove this when we support restarting containers
-			if ctr.state.State == ContainerStateStopped {
-				ctrErrors[id] = errors.Wrapf(ErrNotImplemented, "starting stopped containers is not yet supported")
-
-				delete(ctrsToStart, id)
-				continue
-			}
-
 			// TODO should we only do a dependencies check if we are not ContainerStateCreated?
 			depsOK := true
 			var depErr error