diff --git a/Makefile b/Makefile
index b7029bf4e..40de2576b 100644
--- a/Makefile
+++ b/Makefile
@@ -194,7 +194,7 @@ bin/cri-integration.test:
 
 cri-integration: binaries bin/cri-integration.test ## run cri integration tests
 	@echo "$(WHALE) $@"
-	@./script/test/cri-integration.sh
+	@bash -x ./script/test/cri-integration.sh
 	@rm -rf bin/cri-integration.test
 
 benchmark: ## run benchmarks tests
diff --git a/integration/common.go b/integration/common.go
index 19ea40e9f..3c1e094bc 100644
--- a/integration/common.go
+++ b/integration/common.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
diff --git a/integration/container_log_test.go b/integration/container_log_test.go
index 6c226c13b..e9af33e77 100644
--- a/integration/container_log_test.go
+++ b/integration/container_log_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
diff --git a/integration/container_restart_test.go b/integration/container_restart_test.go
index bce558e88..93e138f3b 100644
--- a/integration/container_restart_test.go
+++ b/integration/container_restart_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -35,6 +33,9 @@ func TestContainerRestart(t *testing.T) {
 		assert.NoError(t, runtimeService.StopPodSandbox(sb))
 		assert.NoError(t, runtimeService.RemovePodSandbox(sb))
 	}()
+
+	EnsureImageExists(t, pauseImage)
+
 	t.Logf("Create a container config and run container in a pod")
 	containerConfig := ContainerConfig(
 		"container1",
diff --git a/integration/container_stats_test.go b/integration/container_stats_test.go
index a3d402217..d75787146 100644
--- a/integration/container_stats_test.go
+++ b/integration/container_stats_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -40,6 +38,9 @@ func TestContainerStats(t *testing.T) {
 		assert.NoError(t, runtimeService.StopPodSandbox(sb))
 		assert.NoError(t, runtimeService.RemovePodSandbox(sb))
 	}()
+
+	EnsureImageExists(t, pauseImage)
+
 	t.Logf("Create a container config and run container in a pod")
 	containerConfig := ContainerConfig(
 		"container1",
@@ -64,8 +65,7 @@ func TestContainerStats(t *testing.T) {
 		if err != nil {
 			return false, err
 		}
-		if s.GetWritableLayer().GetUsedBytes().GetValue() != 0 &&
-			s.GetWritableLayer().GetInodesUsed().GetValue() != 0 {
+		if s.GetWritableLayer().GetUsedBytes().GetValue() != 0 {
 			return true, nil
 		}
 		return false, nil
@@ -162,6 +162,9 @@ func TestContainerListStats(t *testing.T) {
 		assert.NoError(t, runtimeService.StopPodSandbox(sb))
 		assert.NoError(t, runtimeService.RemovePodSandbox(sb))
 	}()
+
+	EnsureImageExists(t, pauseImage)
+
 	t.Logf("Create a container config and run containers in a pod")
 	containerConfigMap := make(map[string]*runtime.ContainerConfig)
 	for i := 0; i < 3; i++ {
@@ -192,8 +195,7 @@ func TestContainerListStats(t *testing.T) {
 			return false, err
 		}
 		for _, s := range stats {
-			if s.GetWritableLayer().GetUsedBytes().GetValue() == 0 &&
-				s.GetWritableLayer().GetInodesUsed().GetValue() == 0 {
+			if s.GetWritableLayer().GetUsedBytes().GetValue() == 0 {
 				return false, nil
 			}
 		}
@@ -217,6 +219,9 @@ func TestContainerListStatsWithIdFilter(t *testing.T) {
 		assert.NoError(t, runtimeService.StopPodSandbox(sb))
 		assert.NoError(t, runtimeService.RemovePodSandbox(sb))
 	}()
+
+	EnsureImageExists(t, pauseImage)
+
 	t.Logf("Create a container config and run containers in a pod")
 	containerConfigMap := make(map[string]*runtime.ContainerConfig)
 	for i := 0; i < 3; i++ {
@@ -251,8 +256,7 @@ func TestContainerListStatsWithIdFilter(t *testing.T) {
 			if len(stats) != 1 {
 				return false, errors.New("unexpected stats length")
 			}
-			if stats[0].GetWritableLayer().GetUsedBytes().GetValue() != 0 &&
-				stats[0].GetWritableLayer().GetInodesUsed().GetValue() != 0 {
+			if stats[0].GetWritableLayer().GetUsedBytes().GetValue() != 0 {
 				return true, nil
 			}
 			return false, nil
@@ -277,6 +281,9 @@ func TestContainerListStatsWithSandboxIdFilter(t *testing.T) {
 		assert.NoError(t, runtimeService.StopPodSandbox(sb))
 		assert.NoError(t, runtimeService.RemovePodSandbox(sb))
 	}()
+
+	EnsureImageExists(t, pauseImage)
+
 	t.Logf("Create a container config and run containers in a pod")
 	containerConfigMap := make(map[string]*runtime.ContainerConfig)
 	for i := 0; i < 3; i++ {
@@ -310,12 +317,12 @@ func TestContainerListStatsWithSandboxIdFilter(t *testing.T) {
 		if len(stats) != 3 {
 			return false, errors.New("unexpected stats length")
 		}
-		if stats[0].GetWritableLayer().GetUsedBytes().GetValue() != 0 &&
-			stats[0].GetWritableLayer().GetInodesUsed().GetValue() != 0 {
+		if stats[0].GetWritableLayer().GetUsedBytes().GetValue() != 0 {
 			return true, nil
 		}
 		return false, nil
-	}, time.Second, 30*time.Second))
+	}, time.Second, 45*time.Second))
+	// TODO(claudiub): Reduce the timer above to 30 seconds once Windows flakiness has been addressed.
 	t.Logf("Verify container stats for sandbox %q", sb)
 	for _, s := range stats {
 		testStats(t, s, containerConfigMap[s.GetAttributes().GetId()])
@@ -333,6 +340,9 @@ func TestContainerListStatsWithIdSandboxIdFilter(t *testing.T) {
 		assert.NoError(t, runtimeService.StopPodSandbox(sb))
 		assert.NoError(t, runtimeService.RemovePodSandbox(sb))
 	}()
+
+	EnsureImageExists(t, pauseImage)
+
 	t.Logf("Create container config and run containers in a pod")
 	containerConfigMap := make(map[string]*runtime.ContainerConfig)
 	for i := 0; i < 3; i++ {
@@ -366,8 +376,7 @@ func TestContainerListStatsWithIdSandboxIdFilter(t *testing.T) {
 			if len(stats) != 1 {
 				return false, errors.New("unexpected stats length")
 			}
-			if stats[0].GetWritableLayer().GetUsedBytes().GetValue() != 0 &&
-				stats[0].GetWritableLayer().GetInodesUsed().GetValue() != 0 {
+			if stats[0].GetWritableLayer().GetUsedBytes().GetValue() != 0 {
 				return true, nil
 			}
 			return false, nil
@@ -389,8 +398,7 @@ func TestContainerListStatsWithIdSandboxIdFilter(t *testing.T) {
 			if len(stats) != 1 {
 				return false, errors.New("unexpected stats length")
 			}
-			if stats[0].GetWritableLayer().GetUsedBytes().GetValue() != 0 &&
-				stats[0].GetWritableLayer().GetInodesUsed().GetValue() != 0 {
+			if stats[0].GetWritableLayer().GetUsedBytes().GetValue() != 0 {
 				return true, nil
 			}
 			return false, nil
@@ -421,5 +429,9 @@ func testStats(t *testing.T,
 	require.NotEmpty(t, s.GetWritableLayer().GetTimestamp())
 	require.NotEmpty(t, s.GetWritableLayer().GetFsId().GetMountpoint())
 	require.NotEmpty(t, s.GetWritableLayer().GetUsedBytes().GetValue())
-	require.NotEmpty(t, s.GetWritableLayer().GetInodesUsed().GetValue())
+
+	// Windows does not collect inodes stats.
+	if goruntime.GOOS != "windows" {
+		require.NotEmpty(t, s.GetWritableLayer().GetInodesUsed().GetValue())
+	}
 }
diff --git a/integration/container_stop_test.go b/integration/container_stop_test.go
index fd5d6eb12..c858d0d84 100644
--- a/integration/container_stop_test.go
+++ b/integration/container_stop_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -20,6 +18,7 @@ package integration
 
 import (
 	"context"
+	goruntime "runtime"
 	"testing"
 	"time"
 
@@ -73,6 +72,9 @@ func TestSharedPidMultiProcessContainerStop(t *testing.T) {
 }
 
 func TestContainerStopCancellation(t *testing.T) {
+	if goruntime.GOOS == "windows" {
+		t.Skip("Skipped on Windows.")
+	}
 	t.Log("Create a pod sandbox")
 	sbConfig := PodSandboxConfig("sandbox", "cancel-container-stop")
 	sb, err := runtimeService.RunPodSandbox(sbConfig, *runtimeHandler)
diff --git a/integration/container_update_resources_test.go b/integration/container_update_resources_test.go
index 83850e0b7..a2196b4cc 100644
--- a/integration/container_update_resources_test.go
+++ b/integration/container_update_resources_test.go
@@ -39,6 +39,7 @@ func checkMemoryLimit(t *testing.T, spec *runtimespec.Spec, memLimit int64) {
 }
 
 func TestUpdateContainerResources(t *testing.T) {
+	// TODO(claudiub): Make this test work once https://github.com/microsoft/hcsshim/pull/931 merges.
 	t.Log("Create a sandbox")
 	sbConfig := PodSandboxConfig("sandbox", "update-container-resources")
 	sb, err := runtimeService.RunPodSandbox(sbConfig, *runtimeHandler)
@@ -48,6 +49,8 @@ func TestUpdateContainerResources(t *testing.T) {
 		assert.NoError(t, runtimeService.RemovePodSandbox(sb))
 	}()
 
+	EnsureImageExists(t, pauseImage)
+
 	t.Log("Create a container with memory limit")
 	cnConfig := ContainerConfig(
 		"container",
diff --git a/integration/container_without_image_ref_test.go b/integration/container_without_image_ref_test.go
index 4be07b147..aeb57c993 100644
--- a/integration/container_without_image_ref_test.go
+++ b/integration/container_without_image_ref_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
diff --git a/integration/containerd_image_test.go b/integration/containerd_image_test.go
index 9907a1d36..55b8f1d52 100644
--- a/integration/containerd_image_test.go
+++ b/integration/containerd_image_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -135,7 +133,7 @@ func TestContainerdImage(t *testing.T) {
 	cnConfig := ContainerConfig(
 		"test-container",
 		id,
-		WithCommand("top"),
+		WithCommand("sleep", "300"),
 	)
 	cn, err := runtimeService.CreateContainer(sb, cnConfig, sbConfig)
 	require.NoError(t, err)
diff --git a/integration/duplicate_name_test.go b/integration/duplicate_name_test.go
index 6c8532063..c391c936a 100644
--- a/integration/duplicate_name_test.go
+++ b/integration/duplicate_name_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -39,6 +37,8 @@ func TestDuplicateName(t *testing.T) {
 	_, err = runtimeService.RunPodSandbox(sbConfig, *runtimeHandler)
 	require.Error(t, err)
 
+	EnsureImageExists(t, pauseImage)
+
 	t.Logf("Create a container")
 	cnConfig := ContainerConfig(
 		"container",
diff --git a/integration/image_load_test.go b/integration/image_load_test.go
index 6de92d49e..db3371483 100644
--- a/integration/image_load_test.go
+++ b/integration/image_load_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -22,6 +20,7 @@ import (
 	"io/ioutil"
 	"os"
 	"os/exec"
+	"path/filepath"
 	"testing"
 	"time"
 
@@ -41,11 +40,12 @@ func TestImageLoad(t *testing.T) {
 	t.Logf("docker save image into tarball")
 	output, err := exec.Command("docker", "pull", testImage).CombinedOutput()
 	require.NoError(t, err, "output: %q", output)
-	tarF, err := ioutil.TempFile("", "image-load")
-	tar := tarF.Name()
+	// ioutil.TempFile also opens a file, which might prevent us from overwriting that file with docker save.
+	tarDir, err := ioutil.TempDir("", "image-load")
+	tar := filepath.Join(tarDir, "image.tar")
 	require.NoError(t, err)
 	defer func() {
-		assert.NoError(t, os.RemoveAll(tar))
+		assert.NoError(t, os.RemoveAll(tarDir))
 	}()
 	output, err = exec.Command("docker", "save", testImage, "-o", tar).CombinedOutput()
 	require.NoError(t, err, "output: %q", output)
diff --git a/integration/imagefs_info_test.go b/integration/imagefs_info_test.go
index dc9210c10..a79a70ee5 100644
--- a/integration/imagefs_info_test.go
+++ b/integration/imagefs_info_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -61,7 +59,6 @@ func TestImageFSInfo(t *testing.T) {
 		info = stats[0]
 		if info.GetTimestamp() != 0 &&
 			info.GetUsedBytes().GetValue() != 0 &&
-			info.GetInodesUsed().GetValue() != 0 &&
 			info.GetFsId().GetMountpoint() != "" {
 			return true, nil
 		}
diff --git a/integration/main_test.go b/integration/main_test.go
index 33e10541f..cfd3d7fc2 100644
--- a/integration/main_test.go
+++ b/integration/main_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -97,6 +95,7 @@ func ConnectDaemons() error {
 	}
 	// containerdEndpoint is the same with criEndpoint now
 	containerdEndpoint = strings.TrimPrefix(*criEndpoint, "unix://")
+	containerdEndpoint = strings.TrimPrefix(containerdEndpoint, "npipe:")
 	containerdClient, err = containerd.New(containerdEndpoint, containerd.WithDefaultNamespace(k8sNamespace))
 	if err != nil {
 		return errors.Wrap(err, "failed to connect containerd")
@@ -198,7 +197,7 @@ func WithTestAnnotations() ContainerOpts {
 }
 
 // Add container resource limits.
-func WithResources(r *runtime.LinuxContainerResources) ContainerOpts {
+func WithResources(r *runtime.LinuxContainerResources) ContainerOpts { //nolint:unused
 	return func(c *runtime.ContainerConfig) {
 		if c.Linux == nil {
 			c.Linux = &runtime.LinuxContainerConfig{}
@@ -240,7 +239,7 @@ func WithLogPath(path string) ContainerOpts {
 }
 
 // WithSupplementalGroups adds supplemental groups.
-func WithSupplementalGroups(gids []int64) ContainerOpts {
+func WithSupplementalGroups(gids []int64) ContainerOpts { //nolint:unused
 	return func(c *runtime.ContainerConfig) {
 		if c.Linux == nil {
 			c.Linux = &runtime.LinuxContainerConfig{}
@@ -325,7 +324,7 @@ func KillProcess(name string) error {
 }
 
 // KillPid kills the process by pid. kill is used.
-func KillPid(pid int) error {
+func KillPid(pid int) error { //nolint:unused
 	output, err := exec.Command("kill", strconv.Itoa(pid)).CombinedOutput()
 	if err != nil {
 		return errors.Errorf("failed to kill %d - error: %v, output: %q", pid, err, output)
@@ -379,7 +378,7 @@ func CRIConfig() (*criconfig.Config, error) {
 }
 
 // SandboxInfo gets sandbox info.
-func SandboxInfo(id string) (*runtime.PodSandboxStatus, *server.SandboxInfo, error) {
+func SandboxInfo(id string) (*runtime.PodSandboxStatus, *server.SandboxInfo, error) { //nolint:unused
 	client, err := RawRuntimeClient()
 	if err != nil {
 		return nil, nil, errors.Wrap(err, "failed to get raw runtime client")
diff --git a/integration/no_metadata_test.go b/integration/no_metadata_test.go
index 5766e19b4..33037b659 100644
--- a/integration/no_metadata_test.go
+++ b/integration/no_metadata_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
diff --git a/integration/pod_dualstack_test.go b/integration/pod_dualstack_test.go
index 339460ab8..b5d31d427 100644
--- a/integration/pod_dualstack_test.go
+++ b/integration/pod_dualstack_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -24,6 +22,7 @@ import (
 	"os"
 	"path/filepath"
 	"regexp"
+	goruntime "runtime"
 	"testing"
 	"time"
 
@@ -54,10 +53,16 @@ func TestPodDualStack(t *testing.T) {
 	EnsureImageExists(t, testImage)
 
 	t.Log("Create a container to print env")
+	var command ContainerOpts
+	if goruntime.GOOS == "windows" {
+		command = WithCommand("ipconfig")
+	} else {
+		command = WithCommand("ip", "address", "show", "dev", "eth0")
+	}
 	cnConfig := ContainerConfig(
 		containerName,
 		testImage,
-		WithCommand("ip", "address", "show", "dev", "eth0"),
+		command,
 		WithLogPath(containerName),
 	)
 	cn, err := runtimeService.CreateContainer(sb, cnConfig, sbConfig)
@@ -85,9 +90,18 @@ func TestPodDualStack(t *testing.T) {
 	ip := status.GetNetwork().GetIp()
 	additionalIps := status.GetNetwork().GetAdditionalIps()
 
-	ipv4Enabled, err := regexp.MatchString("inet .* scope global", string(content))
+	var ipv4Regex, ipv6Regex string
+	if goruntime.GOOS == "windows" {
+		ipv4Regex = "^\\s*IPv4 Address"
+		ipv6Regex = "^\\s*IPv6 Address"
+	} else {
+		ipv4Regex = "inet .* scope global"
+		ipv6Regex = "inet6 .* scope global"
+	}
+
+	ipv4Enabled, err := regexp.MatchString(ipv4Regex, string(content))
 	assert.NoError(t, err)
-	ipv6Enabled, err := regexp.MatchString("inet6 .* scope global", string(content))
+	ipv6Enabled, err := regexp.MatchString(ipv6Regex, string(content))
 	assert.NoError(t, err)
 
 	if ipv4Enabled && ipv6Enabled {
diff --git a/integration/pod_hostname_test.go b/integration/pod_hostname_test.go
index dcebee2b5..c13f857b8 100644
--- a/integration/pod_hostname_test.go
+++ b/integration/pod_hostname_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -22,6 +20,8 @@ import (
 	"io/ioutil"
 	"os"
 	"path/filepath"
+	goruntime "runtime"
+	"strings"
 	"testing"
 	"time"
 
@@ -37,6 +37,7 @@ func TestPodHostname(t *testing.T) {
 		opts             []PodSandboxOpts
 		expectedHostname string
 		expectErr        bool
+		needsHostNetwork bool
 	}{
 		"regular pod with custom hostname": {
 			opts: []PodSandboxOpts{
@@ -49,16 +50,21 @@ func TestPodHostname(t *testing.T) {
 				WithHostNetwork,
 			},
 			expectedHostname: hostname,
+			needsHostNetwork: true,
 		},
 		"host network pod with custom hostname should fail": {
 			opts: []PodSandboxOpts{
 				WithHostNetwork,
 				WithPodHostname("test-hostname"),
 			},
-			expectErr: true,
+			expectErr:        true,
+			needsHostNetwork: true,
 		},
 	} {
 		t.Run(name, func(t *testing.T) {
+			if test.needsHostNetwork && goruntime.GOOS == "windows" {
+				t.Skip("Skipped on Windows.")
+			}
 			testPodLogDir, err := ioutil.TempDir("/tmp", "hostname")
 			require.NoError(t, err)
 			defer os.RemoveAll(testPodLogDir)
@@ -94,7 +100,7 @@ func TestPodHostname(t *testing.T) {
 				containerName,
 				testImage,
 				WithCommand("sh", "-c",
-					"echo -n /etc/hostname= && cat /etc/hostname && env"),
+					"echo -n /etc/hostname= && hostname && env"),
 				WithLogPath(containerName),
 			)
 			cn, err := runtimeService.CreateContainer(sb, cnConfig, sbConfig)
@@ -119,7 +125,11 @@ func TestPodHostname(t *testing.T) {
 			assert.NoError(t, err)
 
 			t.Log("Search hostname env in container log")
-			assert.Contains(t, string(content), "HOSTNAME="+test.expectedHostname)
+			if goruntime.GOOS == "windows" {
+				assert.Contains(t, string(content), "COMPUTERNAME="+strings.ToUpper(test.expectedHostname))
+			} else {
+				assert.Contains(t, string(content), "HOSTNAME="+test.expectedHostname)
+			}
 
 			t.Log("Search /etc/hostname content in container log")
 			assert.Contains(t, string(content), "/etc/hostname="+test.expectedHostname)
diff --git a/integration/restart_test.go b/integration/restart_test.go
index 38dc8693e..c562b1773 100644
--- a/integration/restart_test.go
+++ b/integration/restart_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -19,6 +17,7 @@
 package integration
 
 import (
+	goruntime "runtime"
 	"sort"
 	"testing"
 
@@ -33,6 +32,9 @@ import (
 // Restart test must run sequentially.
 
 func TestContainerdRestart(t *testing.T) {
+	if goruntime.GOOS == "windows" {
+		t.Skip("Skipped on Windows.")
+	}
 	type container struct {
 		name  string
 		id    string
@@ -100,6 +102,9 @@ func TestContainerdRestart(t *testing.T) {
 			runtimeService.StopPodSandbox(sid)
 			runtimeService.RemovePodSandbox(sid)
 		}()
+
+		EnsureImageExists(t, pauseImage)
+
 		s.id = sid
 		for j := range s.containers {
 			c := &s.containers[j]
diff --git a/integration/truncindex_test.go b/integration/truncindex_test.go
index be9ebc9d6..52226d493 100644
--- a/integration/truncindex_test.go
+++ b/integration/truncindex_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -19,6 +17,7 @@
 package integration
 
 import (
+	goruntime "runtime"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
@@ -82,7 +81,7 @@ func TestTruncIndex(t *testing.T) {
 	cnConfig := ContainerConfig(
 		"containerTruncIndex",
 		appImage,
-		WithCommand("top"),
+		WithCommand("sleep", "300"),
 	)
 	cn, err := runtimeService.CreateContainer(sbTruncIndex, cnConfig, sbConfig)
 	require.NoError(t, err)
@@ -112,11 +111,15 @@ func TestTruncIndex(t *testing.T) {
 	require.NoError(t, err)
 	assert.Equal(t, cn, cStats.Attributes.Id)
 
-	t.Logf("Update container memory limit after started")
-	err = runtimeService.UpdateContainerResources(cnTruncIndex, &runtimeapi.LinuxContainerResources{
-		MemoryLimitInBytes: 50 * 1024 * 1024,
-	})
-	assert.NoError(t, err)
+	if goruntime.GOOS != "windows" {
+		// TODO(claudiub): remove this when UpdateContainerResources works on running Windows Containers.
+		// https://github.com/containerd/containerd/issues/5187
+		t.Logf("Update container memory limit after started")
+		err = runtimeService.UpdateContainerResources(cnTruncIndex, &runtimeapi.LinuxContainerResources{
+			MemoryLimitInBytes: 50 * 1024 * 1024,
+		})
+		assert.NoError(t, err)
+	}
 
 	t.Logf("Execute cmd in container")
 	execReq := &runtimeapi.ExecRequest{
diff --git a/integration/volume_copy_up_test.go b/integration/volume_copy_up_test.go
index c03212772..be3208305 100644
--- a/integration/volume_copy_up_test.go
+++ b/integration/volume_copy_up_test.go
@@ -1,5 +1,3 @@
-// +build linux
-
 /*
    Copyright The containerd Authors.
 
@@ -21,6 +19,7 @@ package integration
 import (
 	"fmt"
 	"os/exec"
+	goruntime "runtime"
 	"testing"
 	"time"
 
@@ -29,6 +28,11 @@ import (
 )
 
 func TestVolumeCopyUp(t *testing.T) {
+	if goruntime.GOOS == "windows" {
+		// TODO(claudiub): Remove this when the volume-copy-up image has Windows support.
+		// https://github.com/containerd/containerd/pull/5162
+		t.Skip("Skipped on Windows.")
+	}
 	var (
 		testImage   = GetImage(VolumeCopyUp)
 		execTimeout = time.Minute
@@ -89,6 +93,9 @@ func TestVolumeCopyUp(t *testing.T) {
 }
 
 func TestVolumeOwnership(t *testing.T) {
+	if goruntime.GOOS == "windows" {
+		t.Skip("Skipped on Windows.")
+	}
 	var (
 		testImage   = GetImage(VolumeOwnership)
 		execTimeout = time.Minute
diff --git a/script/test/cri-integration.sh b/script/test/cri-integration.sh
index 795ff8ec3..1aaac0153 100755
--- a/script/test/cri-integration.sh
+++ b/script/test/cri-integration.sh
@@ -33,7 +33,7 @@ mkdir -p ${REPORT_DIR}
 test_setup ${REPORT_DIR}
 
 # Run integration test.
-sudo PATH=${PATH} bin/cri-integration.test --test.run="${FOCUS}" --test.v \
+${sudo} bin/cri-integration.test --test.run="${FOCUS}" --test.v \
   --cri-endpoint=${CONTAINERD_SOCK} \
   --cri-root=${CRI_ROOT} \
   --runtime-handler=${RUNTIME} \
diff --git a/script/test/utils.sh b/script/test/utils.sh
index 6093bf8be..6a40d324d 100755
--- a/script/test/utils.sh
+++ b/script/test/utils.sh
@@ -16,6 +16,11 @@
 
 ROOT="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"/../..
 
+IS_WINDOWS=0
+if [ -v "OS" ] && [ "${OS}" == "Windows_NT" ]; then
+  IS_WINDOWS=1
+fi
+
 # RESTART_WAIT_PERIOD is the period to wait before restarting containerd.
 RESTART_WAIT_PERIOD=${RESTART_WAIT_PERIOD:-10}
 # CONTAINERD_FLAGS contains all containerd flags.
@@ -52,17 +57,35 @@ CONTAINERD_ROOT=${CONTAINERD_ROOT:-"/var/lib/containerd${CONTAINERD_TEST_SUFFIX}
 # The containerd state directory.
 CONTAINERD_STATE=${CONTAINERD_STATE:-"/run/containerd${CONTAINERD_TEST_SUFFIX}"}
 # The containerd socket address.
-CONTAINERD_SOCK=${CONTAINERD_SOCK:-unix://${CONTAINERD_STATE}/containerd.sock}
+if [ $IS_WINDOWS -eq 0 ]; then
+  CONTAINERD_SOCK=${CONTAINERD_SOCK:-unix://${CONTAINERD_STATE}/containerd.sock}
+  TRIMMED_CONTAINERD_SOCK="${CONTAINERD_SOCK#unix://}"
+else
+  CONTAINERD_SOCK=${CONTAINERD_SOCK:-npipe://./pipe/${CONTAINERD_STATE}/containerd}
+  TRIMMED_CONTAINERD_SOCK="${CONTAINERD_SOCK#npipe:}"
+fi
+
 # The containerd binary name.
-CONTAINERD_BIN=${CONTAINERD_BIN:-"containerd"} # don't need a suffix now
+EXE_SUFFIX=""
+if [ $IS_WINDOWS -eq 1 ]; then
+  EXE_SUFFIX=".exe"
+fi
+CONTAINERD_BIN=${CONTAINERD_BIN:-"containerd"}${EXE_SUFFIX}
 if [ -f "${CONTAINERD_CONFIG_FILE}" ]; then
   CONTAINERD_FLAGS+="--config ${CONTAINERD_CONFIG_FILE} "
 fi
-CONTAINERD_FLAGS+="--address ${CONTAINERD_SOCK#"unix://"} \
+
+CONTAINERD_FLAGS+="--address ${TRIMMED_CONTAINERD_SOCK} \
   --state ${CONTAINERD_STATE} \
   --root ${CONTAINERD_ROOT}"
 
-containerd_groupid=
+kill_containerd_cmd=
+
+# NOTE: We don't have the sudo command on Windows.
+sudo=""
+if [ $(id -u) -ne 0 ] && command -v sudo &> /dev/null; then
+  sudo="sudo PATH=${PATH}"
+fi
 
 # test_setup starts containerd.
 test_setup() {
@@ -75,11 +98,17 @@ test_setup() {
   set -m
   # Create containerd in a different process group
   # so that we can easily clean them up.
-  keepalive "sudo PATH=${PATH} bin/containerd ${CONTAINERD_FLAGS}" \
+  keepalive "${sudo} bin/containerd ${CONTAINERD_FLAGS}" \
     ${RESTART_WAIT_PERIOD} &> ${report_dir}/containerd.log &
   pid=$!
   set +m
-  containerd_groupid=$(ps -o pgid= -p ${pid})
+
+  if [ $IS_WINDOWS -eq 1 ]; then
+    kill_containerd_cmd="${sudo} kill ${pid}"
+  else
+    kill_containerd_cmd="${sudo} pkill -g $(ps -o pgid= -p ${pid})"
+  fi
+
   # Wait for containerd to be running by using the containerd client ctr to check the version
   # of the containerd server. Wait an increasing amount of time after each of five attempts
   local -r crictl_path=$(which crictl)
@@ -87,14 +116,14 @@ test_setup() {
     echo "crictl is not in PATH"
     exit 1
   fi
-  readiness_check "sudo bin/ctr --address ${CONTAINERD_SOCK#"unix://"} version"
-  readiness_check "sudo ${crictl_path} --runtime-endpoint=${CONTAINERD_SOCK} info"
+  readiness_check "${sudo} bin/ctr --address ${TRIMMED_CONTAINERD_SOCK} version"
+  readiness_check "${sudo} ${crictl_path} --runtime-endpoint=${CONTAINERD_SOCK} info"
 }
 
 # test_teardown kills containerd.
 test_teardown() {
-  if [ -n "${containerd_groupid}" ]; then
-    sudo pkill -g ${containerd_groupid}
+  if [ -n "${kill_containerd_cmd}" ]; then
+    ${kill_containerd_cmd}
   fi
 }