diff --git a/images/base/files/etc/systemd/system/kubelet.service.d/10-kubeadm.conf b/images/base/files/etc/systemd/system/kubelet.service.d/10-kubeadm.conf
index 17d6239f34..04fd985301 100644
--- a/images/base/files/etc/systemd/system/kubelet.service.d/10-kubeadm.conf
+++ b/images/base/files/etc/systemd/system/kubelet.service.d/10-kubeadm.conf
@@ -8,5 +8,8 @@ EnvironmentFile=-/var/lib/kubelet/kubeadm-flags.env
 # This is a file that the user can use for overrides of the kubelet args as a last resort. Preferably, the user should use
 # the .NodeRegistration.KubeletExtraArgs object in the configuration files instead. KUBELET_EXTRA_ARGS should be sourced from this file.
 EnvironmentFile=-/etc/default/kubelet
+# On cgroup v1, the /kubelet cgroup is created in the entrypoint script before running systemd.
+# On cgroup v2, the /kubelet cgroup is created here. (See the comments in the entrypoint script for the reason.)
+ExecStartPre=/bin/sh -euc "if [ -f /sys/fs/cgroup/cgroup.controllers ]; then create-kubelet-cgroup-v2; fi"
 ExecStart=
 ExecStart=/usr/bin/kubelet $KUBELET_KUBECONFIG_ARGS $KUBELET_CONFIG_ARGS $KUBELET_KUBEADM_ARGS $KUBELET_EXTRA_ARGS --cgroup-root=/kubelet
diff --git a/images/base/files/usr/local/bin/create-kubelet-cgroup-v2 b/images/base/files/usr/local/bin/create-kubelet-cgroup-v2
new file mode 100755
index 0000000000..3c4ddfe44d
--- /dev/null
+++ b/images/base/files/usr/local/bin/create-kubelet-cgroup-v2
@@ -0,0 +1,46 @@
+#!/bin/bash
+
+# Copyright 2021 The Kubernetes Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+set -o errexit
+set -o nounset
+set -o pipefail
+if [[ ! -f "/sys/fs/cgroup/cgroup.controllers" ]]; then
+	echo 'ERROR: this script should not be called on cgroup v1 hosts' >&2
+	exit 1
+fi
+
+# NOTE: we can't use `test -s` because cgroup.procs is not a regular file.
+if [[ -n "$(cat /sys/fs/cgroup/cgroup.procs)" ]]; then
+	echo 'ERROR: this script needs /sys/fs/cgroup/cgroup.procs to be empty (for writing the top-level cgroup.subtree_control)' >&2
+	# So, this script needs to be called after launching systemd.
+	# This script cannot be called from /usr/local/bin/entrypoint.
+	exit 1
+fi
+
+ensure_subtree_control() {
+	local group=$1
+	# When cgroup.controllers is like "cpu cpuset memory io pids",
+	# cgroup.subtree_control is written with "+cpu +cpuset +memory +io +pids" .
+	sed -e 's/ / +/g' -e 's/^/+/' <"/sys/fs/cgroup/$group/cgroup.controllers" >"/sys/fs/cgroup/$group/cgroup.subtree_control"
+}
+
+# kubelet requires all the controllers (including hugetlb) in /sys/fs/cgroup/cgroup.controllers to be available in
+# /sys/fs/cgroup/kubelet/cgroup.subtree_control.
+#
+# We need to update the top-level cgroup.subtree_controllers as well, because hugetlb is not present in the file by default.
+ensure_subtree_control /
+mkdir -p /sys/fs/cgroup/kubelet
+ensure_subtree_control /kubelet
diff --git a/images/base/files/usr/local/bin/entrypoint b/images/base/files/usr/local/bin/entrypoint
index 71bac034ee..4f1ac63cc2 100755
--- a/images/base/files/usr/local/bin/entrypoint
+++ b/images/base/files/usr/local/bin/entrypoint
@@ -83,6 +83,26 @@ mount_kubelet_cgroup_root() {
 }
 
 fix_cgroup() {
+  if [[ -f "/sys/fs/cgroup/cgroup.controllers" ]]; then
+    echo 'INFO: detected cgroup v2'
+    # Both Docker and Podman enable CgroupNS on cgroup v2 hosts by default.
+    #
+    # So mostly we do not need to mess around with the cgroup path stuff,
+    # however, we still need to create the "/kubelet" cgroup at least.
+    # (Otherwise kubelet fails with `cgroup-root ["kubelet"] doesn't exist` error, see #1969)
+    #
+    # The "/kubelet" cgroup is created in ExecStartPre of the kubeadm service.
+    #
+    # [FAQ: Why not create "/kubelet" cgroup here?]
+    # We can't create the cgroup with controllers here, because /sys/fs/cgroup/cgroup.subtree_control is empty.
+    # And yet we can't write controllers to /sys/fs/cgroup/cgroup.subtree_control by ourselves either, because
+    # /sys/fs/cgroup/cgroup.procs is not empty at this moment.
+    #
+    # After switching from this entrypoint script to systemd, systemd evacuates the processes in the root
+    # group to "/init.scope" group, so we can write the root subtree_control and create "/kubelet" cgroup.
+    return
+  fi
+  echo 'INFO: detected cgroup v1'
   echo 'INFO: fix cgroup mounts for all subsystems'
   # see: https://d2iq.com/blog/running-kind-inside-a-kubernetes-cluster-for-continuous-integration
   # capture initial state before modifying