Merge pull request #351 from woshijpf/fix-no-kata-agent

virtcontainers: process the case that kata-agent doesn't start in VM

virtcontainers/api.go

@@ -44,6 +44,8 @@ func CreateSandbox(sandboxConfig SandboxConfig, factory Factory) (VCSandbox, err
 }
 
 func createSandboxFromConfig(sandboxConfig SandboxConfig, factory Factory) (*Sandbox, error) {
+	var err error
+
 	// Create the sandbox.
 	s, err := createSandbox(sandboxConfig, factory)
 	if err != nil {
@@ -51,22 +53,50 @@ func createSandboxFromConfig(sandboxConfig SandboxConfig, factory Factory) (*San
 	}
 
 	// Create the sandbox network
-	if err := s.createNetwork(); err != nil {
+	if err = s.createNetwork(); err != nil {
 		return nil, err
 	}
 
+	// network rollback
+	defer func() {
+		if err != nil && s.networkNS.NetNsCreated {
+			s.removeNetwork()
+		}
+	}()
+
 	// Start the VM
-	if err := s.startVM(); err != nil {
+	if err = s.startVM(); err != nil {
 		return nil, err
 	}
 
+	// rollback to stop VM if error occurs
+	defer func() {
+		if err != nil {
+			s.stopVM()
+		}
+	}()
+
+	// Once startVM is done, we want to guarantee
+	// that the sandbox is manageable. For that we need
+	// to start the sandbox inside the VM.
+	if err = s.agent.startSandbox(s); err != nil {
+		return nil, err
+	}
+
+	// rollback to stop sandbox in VM
+	defer func() {
+		if err != nil {
+			s.agent.stopSandbox(s)
+		}
+	}()
+
 	// Create Containers
-	if err := s.createContainers(); err != nil {
+	if err = s.createContainers(); err != nil {
 		return nil, err
 	}
 
 	// The sandbox is completely created now, we can store it.
-	if err := s.storeSandbox(); err != nil {
+	if err = s.storeSandbox(); err != nil {
 		return nil, err
 	}
 

virtcontainers/kata_agent.go

@@ -450,6 +450,8 @@ func (k *kataAgent) generateInterfacesAndRoutes(networkNS NetworkNamespace) ([]*
 }
 
 func (k *kataAgent) startProxy(sandbox *Sandbox) error {
+	var err error
+
 	if k.proxy == nil {
 		return errorMissingProxy
 	}
@@ -475,10 +477,18 @@ func (k *kataAgent) startProxy(sandbox *Sandbox) error {
 		return err
 	}
 
+	// If error occurs after kata-proxy process start,
+	// then rollback to kill kata-proxy process
+	defer func() {
+		if err != nil && pid > 0 {
+			k.proxy.stop(sandbox, pid)
+		}
+	}()
+
 	// Fill agent state with proxy information, and store them.
 	k.state.ProxyPid = pid
 	k.state.URL = uri
-	if err := sandbox.storage.storeAgentState(sandbox.id, k.state); err != nil {
+	if err = sandbox.storage.storeAgentState(sandbox.id, k.state); err != nil {
 		return err
 	}
 
@@ -502,6 +512,11 @@ func (k *kataAgent) startSandbox(sandbox *Sandbox) error {
 		hostname = hostname[:maxHostnameLen]
 	}
 
+	// check grpc server is serving
+	if err = k.check(); err != nil {
+		return err
+	}
+
 	//
 	// Setup network interfaces and routes
 	//

virtcontainers/sandbox.go

@@ -988,10 +988,12 @@ func (s *Sandbox) startVM() error {
 
 	s.Logger().Info("VM started")
 
-	// Once startVM is done, we want to guarantee
-	// that the sandbox is manageable. For that we need
-	// to start the sandbox inside the VM.
-	return s.agent.startSandbox(s)
+	return nil
+}
+
+// stopVM: stop the sandbox's VM
+func (s *Sandbox) stopVM() error {
+	return s.hypervisor.stopSandbox()
 }
 
 func (s *Sandbox) addContainer(c *Container) error {