Bug 1943700

Description of problem:
=======================
We're seeing reports of Rancher installs failing with updated Red Hat packages of Docker.  It appears to be stuck launching the kubelet on the node, and I can easily reproduce it.

This is mentioned here, as well:

https://github.com/rancher/rke/issues/2401

I'm debugging it myself currently as well.

Version-Release number of selected component (if applicable):
=============================================================
# uname -a
Linux rhel7 3.10.0-1160.el7.x86_64 #1 SMP Tue Aug 18 14:50:17 EDT 2020 x86_64 x86_64 x86_64 GNU/Linux

# rpm -q docker
docker-1.13.1-204.git0be3e21.el7_9.x86_64

(I also confirmed on docker-1.13.1-203)

How reproducible:
=================
Every time.

Steps to Reproduce:
===================
Download rke v1.2.4-rc9:

# wget https://github.com/rancher/rke/releases/download/v1.2.4-rc9/rke_linux-amd64 -O /bin/rke

# chmod +x /bin/rke

# rke --version
rke version v1.2.4-rc9

Create an "rke" user and add them to the docker group with a password.

# useradd rke

# usermod -a -G docker rke

# passwd rke

Become the "rke" user and ensure you can run Docker commands.

# sudo su - rke

$ docker ps -a

Generate SSH keys for the "rke" user and copy them:

$ ssh-keygen -t rsa

$ ssh-copy-id rke@localhost

Run `rke config`, I used these values and it will automatically create a `cluster.yml' file:

$ rke config
WARN[0000] This is not an officially supported version (v1.2.4-rc9) of RKE. Please download the latest official release at https://github.com/rancher/rke/releases
[+] Cluster Level SSH Private Key Path [~/.ssh/id_rsa]:
[+] Number of Hosts [1]:
[+] SSH Address of host (1) [none]: localhost
[+] SSH Port of host (1) [22]:
[+] SSH Private Key Path of host (localhost) [none]:
[-] You have entered empty SSH key path, trying fetch from SSH key parameter
[+] SSH Private Key of host (localhost) [none]:
[-] You have entered empty SSH key, defaulting to cluster level SSH key: ~/.ssh/id_rsa
[+] SSH User of host (localhost) [ubuntu]: rke
[+] Is host (localhost) a Control Plane host (y/n)? [y]:
[+] Is host (localhost) a Worker host (y/n)? [n]: y
[+] Is host (localhost) an etcd host (y/n)? [n]: y
[+] Override Hostname of host (localhost) [none]:
[+] Internal IP of host (localhost) [none]:
[+] Docker socket path on host (localhost) [/var/run/docker.sock]:
[+] Network Plugin Type (flannel, calico, weave, canal, aci) [canal]: flannel
[+] Authentication Strategy [x509]:
[+] Authorization Mode (rbac, none) [rbac]: none
[+] Kubernetes Docker image [rancher/hyperkube:v1.19.6-rancher1]:
[+] Cluster domain [cluster.local]:
[+] Service Cluster IP Range [10.43.0.0/16]:
[+] Enable PodSecurityPolicy [n]:
[+] Cluster Network CIDR [10.42.0.0/16]:
[+] Cluster DNS Service IP [10.43.0.10]:
[+] Add addon manifest URLs or YAML files [no]:

Bring the cluster online:

$ rke up
WARN[0000] This is not an officially supported version (v1.2.4-rc9) of RKE. Please download the latest official release at https://github.com/rancher/rke/releases
INFO[0000] Running RKE version: v1.2.4-rc9
INFO[0000] Initiating Kubernetes cluster
- - - - 8< - - - -
INFO[0140] Image [rancher/hyperkube:v1.19.6-rancher1] exists on host [localhost]
INFO[0140] Starting container [kubelet] on host [localhost], try #1
WARN[0190] Can't start Docker container [kubelet] on host [localhost]: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?
INFO[0190] Starting container [kubelet] on host [localhost], try #2
WARN[0240] Can't start Docker container [kubelet] on host [localhost]: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?
INFO[0240] Starting container [kubelet] on host [localhost], try #3

Observe that docker is hung during this step:

[root@rhel7 ~]# docker ps -a
(hung forever)

Actual results:
Docker is frozen when `kubelet` hyperkube rancher container is coming online.

Expected results:
Docker should not freeze/hang.

Additional info:
Downgrading to docker-1.13.1-162.git64e9980.el7_8 resolves the issue.