Merge pull request #47850 from dcbw/checkpoint-hostnetwork

Automatic merge from submit-queue (batch tested with PRs 47850, 47835, 46197, 47250, 48284)

dockershim: checkpoint HostNetwork property

To ensure kubelet doesn't attempt network teardown on HostNetwork
containers that no longer exist but are still checkpointed, make
sure we preserve the HostNetwork property in checkpoints.  If
the checkpoint indicates the container was a HostNetwork one,
don't tear down the network since that would fail anyway.

Related: https://github.com/kubernetes/kubernetes/issues/44307#issuecomment-299548609

@freehan @kubernetes/sig-network-misc

pkg/kubelet/dockershim/docker_checkpoint.go

@@ -50,6 +50,7 @@ type PortMapping struct {
 // CheckpointData contains all types of data that can be stored in the checkpoint.
 type CheckpointData struct {
 	PortMappings []*PortMapping `json:"port_mappings,omitempty"`
+	HostNetwork  bool           `json:"host_network,omitempty"`
 }
 
 // PodSandboxCheckpoint is the checkpoint structure for a sandbox

pkg/kubelet/dockershim/docker_checkpoint_test.go

@@ -48,18 +48,22 @@ func TestPersistentCheckpointHandler(t *testing.T) {
 			&port443,
 		},
 	}
+	checkpoint1.Data.HostNetwork = true
 
 	checkpoints := []struct {
 		podSandboxID      string
 		checkpoint        *PodSandboxCheckpoint
+		expectHostNetwork bool
 	}{
 		{
 			"id1",
 			checkpoint1,
+			true,
 		},
 		{
 			"id2",
 			NewPodSandboxCheckpoint("ns2", "sandbox2"),
+			false,
 		},
 	}
 
@@ -72,6 +76,7 @@ func TestPersistentCheckpointHandler(t *testing.T) {
 		checkpoint, err := handler.GetCheckpoint(tc.podSandboxID)
 		assert.NoError(t, err)
 		assert.Equal(t, *checkpoint, *tc.checkpoint)
+		assert.Equal(t, checkpoint.Data.HostNetwork, tc.expectHostNetwork)
 	}
 	// Test ListCheckpoints
 	keys, err := handler.ListCheckpoints()

pkg/kubelet/dockershim/docker_sandbox.go

@@ -171,14 +171,14 @@ func (ds *dockerService) RunPodSandbox(config *runtimeapi.PodSandboxConfig) (id
 // after us?
 func (ds *dockerService) StopPodSandbox(podSandboxID string) error {
 	var namespace, name string
+	var hostNetwork bool
 	var checkpointErr, statusErr error
-	needNetworkTearDown := false
 
 	// Try to retrieve sandbox information from docker daemon or sandbox checkpoint
 	status, statusErr := ds.PodSandboxStatus(podSandboxID)
 	if statusErr == nil {
 		nsOpts := status.GetLinux().GetNamespaces().GetOptions()
-		needNetworkTearDown = nsOpts != nil && !nsOpts.HostNetwork
+		hostNetwork = nsOpts != nil && nsOpts.HostNetwork
 		m := status.GetMetadata()
 		namespace = m.Namespace
 		name = m.Name
@@ -211,10 +211,8 @@ func (ds *dockerService) StopPodSandbox(podSandboxID string) error {
 		} else {
 			namespace = checkpoint.Namespace
 			name = checkpoint.Name
+			hostNetwork = checkpoint.Data != nil && checkpoint.Data.HostNetwork
 		}
-
-		// Always trigger network plugin to tear down
-		needNetworkTearDown = true
 	}
 
 	// WARNING: The following operations made the following assumption:
@@ -226,7 +224,7 @@ func (ds *dockerService) StopPodSandbox(podSandboxID string) error {
 	// since it is stopped. With empty network namespcae, CNI bridge plugin will conduct best
 	// effort clean up and will not return error.
 	errList := []error{}
-	if needNetworkTearDown {
+	if !hostNetwork {
 		cID := kubecontainer.BuildContainerID(runtimeName, podSandboxID)
 		err := ds.network.TearDownPod(namespace, name, cID)
 		if err == nil {
@@ -642,6 +640,9 @@ func constructPodSandboxCheckpoint(config *runtimeapi.PodSandboxConfig) *PodSand
 			Protocol:      &proto,
 		})
 	}
+	if nsOptions := config.GetLinux().GetSecurityContext().GetNamespaceOptions(); nsOptions != nil {
+		checkpoint.Data.HostNetwork = nsOptions.HostNetwork
+	}
 	return checkpoint
 }