Merge pull request #94 from mheon/restart

Add handling for system restart in libpod

5 files changed, 158 insertions, 5 deletions

diff --git a/libpod/container.go b/libpod/container.go
index 474d4be95..70011963d 100644
--- a/libpod/container.go
+++ b/libpod/container.go
@@ -420,6 +420,30 @@ func (c *Container) teardownStorage() error {
 	return nil
 }
 
+// Refresh refreshes the container's state after a restart
+func (c *Container) refresh() error {
+	c.lock.Lock()
+	defer c.lock.Unlock()
+
+	if !c.valid {
+		return errors.Wrapf(ErrCtrRemoved, "container %s is not valid - may have been removed", c.ID())
+	}
+
+	// We need to get the container's temporary directory from c/storage
+	// It was lost in the reboot and must be recreated
+	dir, err := c.runtime.storageService.GetRunDir(c.ID())
+	if err != nil {
+		return errors.Wrapf(err, "error retrieving temporary directory for container %s", c.ID())
+	}
+	c.state.RunDir = dir
+
+	if err := c.runtime.state.SaveContainer(c); err != nil {
+		return errors.Wrapf(err, "error refreshing state for container %s", c.ID())
+	}
+
+	return nil
+}
+
 // Init creates a container in the OCI runtime
 func (c *Container) Init() (err error) {
 	c.lock.Lock()
@@ -433,12 +457,27 @@ func (c *Container) Init() (err error) {
 		return errors.Wrapf(ErrCtrExists, "container %s has already been created in runtime", c.ID())
 	}
 
-	// Mount storage for the container
 	if err := c.mountStorage(); err != nil {
 		return err
 	}
 
-	// Make the OCI runtime spec we will use
+	// If the OCI spec already exists, we need to replace it
+	// Cannot guarantee some things, e.g. network namespaces, have the same
+	// paths
+	jsonPath := filepath.Join(c.bundlePath(), "config.json")
+	if _, err := os.Stat(jsonPath); err != nil {
+		if !os.IsNotExist(err) {
+			return errors.Wrapf(err, "error doing stat on container %s spec", c.ID())
+		}
+		// The spec does not exist, we're fine
+	} else {
+		// The spec exists, need to remove it
+		if err := os.Remove(jsonPath); err != nil {
+			return errors.Wrapf(err, "error replacing runtime spec for container %s", c.ID())
+		}
+	}
+
+	// Save OCI spec to disk
 	g := generate.NewFromSpec(c.config.Spec)
 	// Mount ShmDir from host into container
 	g.AddBindMount(c.config.ShmDir, "/dev/shm", []string{"rw"})
@@ -447,8 +486,6 @@ func (c *Container) Init() (err error) {
 	c.runningSpec.Annotations[crioAnnotations.Created] = c.config.CreatedTime.Format(time.RFC3339Nano)
 	c.runningSpec.Annotations["org.opencontainers.image.stopSignal"] = fmt.Sprintf("%d", c.config.StopSignal)
 
-	// Save the OCI spec to disk
-	jsonPath := filepath.Join(c.bundlePath(), "config.json")
 	fileJSON, err := json.Marshal(c.runningSpec)
 	if err != nil {
 		return errors.Wrapf(err, "error exporting runtime spec for container %s to JSON", c.ID())
@@ -456,10 +493,11 @@ func (c *Container) Init() (err error) {
 	if err := ioutil.WriteFile(jsonPath, fileJSON, 0644); err != nil {
 		return errors.Wrapf(err, "error writing runtime spec JSON to file for container %s", c.ID())
 	}
-	c.state.ConfigPath = jsonPath
 
 	logrus.Debugf("Created OCI spec for container %s at %s", c.ID(), jsonPath)
 
+	c.state.ConfigPath = jsonPath
+
 	// With the spec complete, do an OCI create
 	// TODO set cgroup parent in a sane fashion
 	if err := c.runtime.ociRuntime.createContainer(c, "/libpod_parent"); err != nil {
diff --git a/libpod/in_memory_state.go b/libpod/in_memory_state.go
index 5d03e62e6..4e4cbb664 100644
--- a/libpod/in_memory_state.go
+++ b/libpod/in_memory_state.go
@@ -38,6 +38,12 @@ func (s *InMemoryState) Close() error {
 	return nil
 }
 
+// Refresh clears container and pod stats after a reboot
+// In-memory state won't survive a reboot so this is a no-op
+func (s *InMemoryState) Refresh() error {
+	return nil
+}
+
 // Container retrieves a container from its full ID
 func (s *InMemoryState) Container(id string) (*Container, error) {
 	if id == "" {
diff --git a/libpod/runtime.go b/libpod/runtime.go
index b86e06546..36225bf69 100644
--- a/libpod/runtime.go
+++ b/libpod/runtime.go
@@ -173,6 +173,36 @@ func NewRuntime(options ...RuntimeOption) (runtime *Runtime, err error) {
 		runtime.state = state
 	}
 
+	// We now need to see if the system has restarted
+	// We check for the presence of a file in our tmp directory to verify this
+	// This check must be locked to prevent races
+	runtimeAliveLock := filepath.Join(runtime.config.TmpDir, "alive.lck")
+	runtimeAliveFile := filepath.Join(runtime.config.TmpDir, "alive")
+	aliveLock, err := storage.GetLockfile(runtimeAliveLock)
+	if err != nil {
+		return nil, errors.Wrapf(err, "error acquiring runtime init lock")
+	}
+	// Acquire the lock and hold it until we return
+	// This ensures that no two processes will be in runtime.refresh at once
+	// TODO: we can't close the FD in this lock, so we should keep it around
+	// and use it to lock important operations
+	aliveLock.Lock()
+	defer aliveLock.Unlock()
+	_, err = os.Stat(runtimeAliveFile)
+	if err != nil {
+		// If the file doesn't exist, we need to refresh the state
+		// This will trigger on first use as well, but refreshing an
+		// empty state only creates a single file
+		// As such, it's not really a performance concern
+		if os.IsNotExist(err) {
+			if err2 := runtime.refresh(runtimeAliveFile); err2 != nil {
+				return nil, err2
+			}
+		} else {
+			return nil, errors.Wrapf(err, "error reading runtime status file %s", runtimeAliveFile)
+		}
+	}
+
 	// Mark the runtime as valid - ready to be used, cannot be modified
 	// further
 	runtime.valid = true
@@ -238,3 +268,33 @@ func (r *Runtime) Shutdown(force bool) error {
 
 	return lastError
 }
+
+// Reconfigures the runtime after a reboot
+// Refreshes the state, recreating temporary files
+// Does not check validity as the runtime is not valid until after this has run
+func (r *Runtime) refresh(alivePath string) error {
+	// First clear the state in the database
+	if err := r.state.Refresh(); err != nil {
+		return err
+	}
+
+	// Next refresh the state of all containers to recreate dirs and
+	// namespaces
+	ctrs, err := r.state.AllContainers()
+	if err != nil {
+		return errors.Wrapf(err, "error retrieving all containers from state")
+	}
+	for _, ctr := range ctrs {
+		if err := ctr.refresh(); err != nil {
+			return err
+		}
+	}
+
+	file, err := os.OpenFile(alivePath, os.O_RDONLY|os.O_CREATE, 0644)
+	if err != nil {
+		return errors.Wrapf(err, "error creating runtime status file %s", alivePath)
+	}
+	defer file.Close()
+
+	return nil
+}
diff --git a/libpod/sql_state.go b/libpod/sql_state.go
index fb6be85d7..bba697d18 100644
--- a/libpod/sql_state.go
+++ b/libpod/sql_state.go
@@ -103,6 +103,52 @@ func (s *SQLState) Close() error {
 	return nil
 }
 
+// Refresh clears the state after a reboot
+// Resets mountpoint, PID, state for all containers
+func (s *SQLState) Refresh() (err error) {
+	const refresh = `UPDATE containerState SET
+                             State=?,
+                             Mountpoint=?,
+                             Pid=?;`
+
+	s.lock.Lock()
+	defer s.lock.Unlock()
+
+	if !s.valid {
+		return ErrDBClosed
+	}
+
+	tx, err := s.db.Begin()
+	if err != nil {
+		return errors.Wrapf(err, "error beginning database transaction")
+	}
+	defer func() {
+		if err != nil {
+			if err2 := tx.Rollback(); err2 != nil {
+				logrus.Errorf("Error rolling back transaction to refresh state: %v", err2)
+			}
+		}
+	}()
+
+	// Refresh container state
+	// The constants could be moved into the SQL, but keeping them here
+	// will keep us in sync in case ContainerStateConfigured ever changes in
+	// the container state
+	_, err = tx.Exec(refresh,
+		ContainerStateConfigured,
+		"",
+		0)
+	if err != nil {
+		return errors.Wrapf(err, "error refreshing database state")
+	}
+
+	if err := tx.Commit(); err != nil {
+		return errors.Wrapf(err, "error committing transaction to refresh database")
+	}
+
+	return nil
+}
+
 // Container retrieves a container from its full ID
 func (s *SQLState) Container(id string) (*Container, error) {
 	const query = `SELECT containers.*,
diff --git a/libpod/state.go b/libpod/state.go
index 4093f14f1..4a79b8d2d 100644
--- a/libpod/state.go
+++ b/libpod/state.go
@@ -6,6 +6,9 @@ type State interface {
 	// connections) that may be required
 	Close() error
 
+	// Refresh clears container and pod states after a reboot
+	Refresh() error
+
 	// Accepts full ID of container
 	Container(id string) (*Container, error)
 	// Accepts full or partial IDs (as long as they are unique) and names