Fix stale sandbox from store problem

At times, when checkpointed sandbox from store cannot be
cleaned up properly we still retain the sandbox in both
the store and in memory. But this sandbox store may not
contain important configuration information from docker.
So when docker requests a new sandbox, instead of using
it as is, reconcile the sandbox state from store with the
the configuration information provided by docker. To do this
mark the sandbox from store as stub and never reveal it to
external searches. When docker requests a new sandbox, update
the stub sandbox and clear the stub flag.

Signed-off-by: Jana Radhakrishnan <mrjana@docker.com>

Author: mrjana

controller.go

@@ -100,6 +100,9 @@ type NetworkController interface {
 	// SandboxByID returns the Sandbox which has the passed id. If not found, a types.NotFoundError is returned.
 	SandboxByID(id string) (Sandbox, error)
 
+	// SandboxDestroy destroys a sandbox given a container ID
+	SandboxDestroy(id string) error
+
 	// Stop network controller
 	Stop()
 }
@@ -478,21 +481,37 @@ func (c *controller) NewSandbox(containerID string, options ...SandboxOption) (S
 		return nil, types.BadRequestErrorf("invalid container ID")
 	}
 
-	var existing Sandbox
-	look := SandboxContainerWalker(&existing, containerID)
-	c.WalkSandboxes(look)
-	if existing != nil {
-		return nil, types.BadRequestErrorf("container %s is already present: %v", containerID, existing)
+	var sb *sandbox
+	c.Lock()
+	for _, s := range c.sandboxes {
+		if s.containerID == containerID {
+			// If not a stub, then we already have a complete sandbox.
+			if !s.isStub {
+				c.Unlock()
+				return nil, types.BadRequestErrorf("container %s is already present: %v", containerID, s)
+			}
+
+			// We already have a stub sandbox from the
+			// store. Make use of it so that we don't lose
+			// the endpoints from store but reset the
+			// isStub flag.
+			sb = s
+			sb.isStub = false
+			break
+		}
 	}
+	c.Unlock()
 
 	// Create sandbox and process options first. Key generation depends on an option
-	sb := &sandbox{
-		id:          stringid.GenerateRandomID(),
-		containerID: containerID,
-		endpoints:   epHeap{},
-		epPriority:  map[string]int{},
-		config:      containerConfig{},
-		controller:  c,
+	if sb == nil {
+		sb = &sandbox{
+			id:          stringid.GenerateRandomID(),
+			containerID: containerID,
+			endpoints:   epHeap{},
+			epPriority:  map[string]int{},
+			config:      containerConfig{},
+			controller:  c,
+		}
 	}
 
 	heap.Init(&sb.endpoints)
@@ -547,6 +566,11 @@ func (c *controller) Sandboxes() []Sandbox {
 
 	list := make([]Sandbox, 0, len(c.sandboxes))
 	for _, s := range c.sandboxes {
+		// Hide stub sandboxes from libnetwork users
+		if s.isStub {
+			continue
+		}
+
 		list = append(list, s)
 	}
 
@@ -574,6 +598,26 @@ func (c *controller) SandboxByID(id string) (Sandbox, error) {
 	return s, nil
 }
 
+// SandboxDestroy destroys a sandbox given a container ID
+func (c *controller) SandboxDestroy(id string) error {
+	var sb *sandbox
+	c.Lock()
+	for _, s := range c.sandboxes {
+		if s.containerID == id {
+			sb = s
+			break
+		}
+	}
+	c.Unlock()
+
+	// It is not an error if sandbox is not available
+	if sb == nil {
+		return nil
+	}
+
+	return sb.Delete()
+}
+
 // SandboxContainerWalker returns a Sandbox Walker function which looks for an existing Sandbox with the passed containerID
 func SandboxContainerWalker(out *Sandbox, containerID string) SandboxWalker {
 	return func(sb Sandbox) bool {

sandbox.go

@@ -68,6 +68,7 @@ type sandbox struct {
 	joinLeaveDone chan struct{}
 	dbIndex       uint64
 	dbExists      bool
+	isStub        bool
 	inDelete      bool
 	sync.Mutex
 }

sandbox_store.go

@@ -188,6 +188,7 @@ func (c *controller) sandboxCleanup() {
 			endpoints:   epHeap{},
 			epPriority:  map[string]int{},
 			dbIndex:     sbs.dbIndex,
+			isStub:      true,
 			dbExists:    true,
 		}