diff --git a/pkg/kubelet/BUILD b/pkg/kubelet/BUILD
index e8d61bdbe0c..d25639a975e 100644
--- a/pkg/kubelet/BUILD
+++ b/pkg/kubelet/BUILD
@@ -250,13 +250,20 @@ go_test(
         "//staging/src/k8s.io/client-go/util/testing:go_default_library",
         "//staging/src/k8s.io/component-base/featuregate/testing:go_default_library",
         "//staging/src/k8s.io/component-base/version:go_default_library",
-        "//staging/src/k8s.io/cri-api/pkg/apis/runtime/v1alpha2:go_default_library",
         "//vendor/github.com/google/cadvisor/info/v1:go_default_library",
         "//vendor/github.com/google/cadvisor/info/v2:go_default_library",
         "//vendor/github.com/stretchr/testify/assert:go_default_library",
         "//vendor/github.com/stretchr/testify/require:go_default_library",
         "//vendor/k8s.io/utils/mount:go_default_library",
-    ],
+    ] + select({
+        "@io_bazel_rules_go//go/platform:android": [
+            "//staging/src/k8s.io/cri-api/pkg/apis/runtime/v1alpha2:go_default_library",
+        ],
+        "@io_bazel_rules_go//go/platform:linux": [
+            "//staging/src/k8s.io/cri-api/pkg/apis/runtime/v1alpha2:go_default_library",
+        ],
+        "//conditions:default": [],
+    }),
 )
 
 filegroup(
diff --git a/pkg/kubelet/container/runtime.go b/pkg/kubelet/container/runtime.go
index 8a4db8970fb..84fa08b80f5 100644
--- a/pkg/kubelet/container/runtime.go
+++ b/pkg/kubelet/container/runtime.go
@@ -285,6 +285,7 @@ type PodStatus struct {
 	// Status of containers in the pod.
 	ContainerStatuses []*ContainerStatus
 	// Status of the pod sandbox.
+	// Only for kuberuntime now, other runtime may keep it nil.
 	SandboxStatuses []*runtimeapi.PodSandboxStatus
 }
 
diff --git a/pkg/kubelet/cri/remote/fake/fake_runtime.go b/pkg/kubelet/cri/remote/fake/fake_runtime.go
index 339561f9b8e..e49f311aa18 100644
--- a/pkg/kubelet/cri/remote/fake/fake_runtime.go
+++ b/pkg/kubelet/cri/remote/fake/fake_runtime.go
@@ -112,7 +112,7 @@ func (f *RemoteRuntime) StopPodSandbox(ctx context.Context, req *kubeapi.StopPod
 // This call is idempotent, and must not return an error if the sandbox has
 // already been removed.
 func (f *RemoteRuntime) RemovePodSandbox(ctx context.Context, req *kubeapi.RemovePodSandboxRequest) (*kubeapi.RemovePodSandboxResponse, error) {
-	err := f.RuntimeService.RemovePodSandbox(req.PodSandboxId)
+	err := f.RuntimeService.StopPodSandbox(req.PodSandboxId)
 	if err != nil {
 		return nil, err
 	}
diff --git a/pkg/kubelet/kubelet_pods.go b/pkg/kubelet/kubelet_pods.go
index d6f8442cd4f..d70acf1d803 100644
--- a/pkg/kubelet/kubelet_pods.go
+++ b/pkg/kubelet/kubelet_pods.go
@@ -940,16 +940,6 @@ func (kl *Kubelet) PodResourcesAreReclaimed(pod *v1.Pod, status v1.PodStatus) bo
 		klog.V(3).Infof("Pod %q is terminated, but some containers have not been cleaned up: %s", format.Pod(pod), statusStr)
 		return false
 	}
-	// pod's sandboxes should be deleted
-	if len(runtimeStatus.SandboxStatuses) > 0 {
-		var sandboxStr string
-		for _, sandbox := range runtimeStatus.SandboxStatuses {
-			sandboxStr += fmt.Sprintf("%+v ", *sandbox)
-		}
-		klog.V(3).Infof("Pod %q is terminated, but some pod sandboxes have not been cleaned up: %s", format.Pod(pod), sandboxStr)
-		return false
-	}
-
 	if kl.podVolumesExist(pod.UID) && !kl.keepTerminatedPodVolumes {
 		// We shouldn't delete pods whose volumes have not been cleaned up if we are not keeping terminated pod volumes
 		klog.V(3).Infof("Pod %q is terminated, but some volumes have not been cleaned up", format.Pod(pod))
diff --git a/pkg/kubelet/kubelet_pods_test.go b/pkg/kubelet/kubelet_pods_test.go
index e475a186be0..5886f49d6a9 100644
--- a/pkg/kubelet/kubelet_pods_test.go
+++ b/pkg/kubelet/kubelet_pods_test.go
@@ -42,7 +42,6 @@ import (
 	// api.Registry.GroupOrDie(v1.GroupName).GroupVersions[0].String() is changed
 	// to "v1"?
 
-	runtimeapi "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
 	_ "k8s.io/kubernetes/pkg/apis/core/install"
 	"k8s.io/kubernetes/pkg/features"
 	kubecontainer "k8s.io/kubernetes/pkg/kubelet/container"
@@ -2386,70 +2385,3 @@ func TestTruncatePodHostname(t *testing.T) {
 		assert.Equal(t, test.output, output)
 	}
 }
-
-func TestPodResourcesAreReclaimed(t *testing.T) {
-
-	type args struct {
-		pod           *v1.Pod
-		status        v1.PodStatus
-		runtimeStatus kubecontainer.PodStatus
-	}
-	tests := []struct {
-		name string
-		args args
-		want bool
-	}{
-		{
-			"pod with running containers",
-			args{
-				pod: &v1.Pod{},
-				status: v1.PodStatus{
-					ContainerStatuses: []v1.ContainerStatus{
-						runningState("containerA"),
-						runningState("containerB"),
-					},
-				},
-			},
-			false,
-		},
-		{
-			"pod with containers in runtime cache",
-			args{
-				pod:    &v1.Pod{},
-				status: v1.PodStatus{},
-				runtimeStatus: kubecontainer.PodStatus{
-					ContainerStatuses: []*kubecontainer.ContainerStatus{
-						{},
-					},
-				},
-			},
-			false,
-		},
-		{
-			"pod with sandbox present",
-			args{
-				pod:    &v1.Pod{},
-				status: v1.PodStatus{},
-				runtimeStatus: kubecontainer.PodStatus{
-					SandboxStatuses: []*runtimeapi.PodSandboxStatus{
-						{},
-					},
-				},
-			},
-			false,
-		},
-	}
-
-	testKubelet := newTestKubelet(t, false)
-	defer testKubelet.Cleanup()
-	kl := testKubelet.kubelet
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			testKubelet.fakeRuntime.PodStatus = tt.args.runtimeStatus
-			if got := kl.PodResourcesAreReclaimed(tt.args.pod, tt.args.status); got != tt.want {
-				t.Errorf("PodResourcesAreReclaimed() = %v, want %v", got, tt.want)
-			}
-		})
-	}
-}