Skip to content

OS setting

We are going to install some helpful software, set up hostnames, and do some other configuration that we need before we install Kubernetes. Of course, we are going to use Ansible. Don’t get scared its simple 🙂. If you have followed my guides, you should not have any issues.

Set hostname to every node

If you remember, Here, I have set variable var_hostname in file /etc/ansible/hosts for each node, and this little nugget comes into play right now. On your main control node execute the following:

ansible cube -b -m shell -a "hostnamectl set-hostname {{ var_hostname }}"

Notable parameters are “cube”, as mentioned before, it means it will execute commands on every single node. Next, -b is short for –become. What this does is sudo by default, as we should have an ubuntu user with sudo already set up. This is all we need to execute our commands as root.

Next in line, -m shell. we are telling Ansible to use the shell module. This in essence mean we are going to execute commands via shell, and the commands are in -a ” <commands> “.

Last in line is {{ var_hostname }}. This gets replaced for each node with a variable we set up in /etc/ansible/hosts. Setting variables ahead can be an easy way to customize some nodes, while still letting Ansible do all the heavy lifting.

You can check manually if everything worked by logging in to any of the nodes and running:

hostnamectl status
# or from main control node run Ansible command and get result from all nodes 🙂 Isn't Ansible useful ?
ansible cube -b -m shell -a "hostnamectl status | grep hostname"

Remove “snap”

Don’t get me wrong. Ubuntu’s snap is not a bad idea, just useless for us. We are going to remove this feature and save some resources.

# All nodes should be the same so running this "list" should be enough on the main control node
ubuntu@control01:~$ snap list
Name    Version   Rev    Tracking         Publisher   Notes
core18  20200929  1936   latest/stable    canonical✓  base
lxd     4.7       18140  latest/stable/…  canonical✓  -
snapd   2.47.1    9730   latest/stable    canonical✓  snapd

# Remove the snaps from all nodes ( order matters )
ansible cube -b -m shell -a "snap remove lxd && snap remove core18 && snap remove snapd"

Remove snapd demon:

ansible cube -b -m shell -a "apt purge snapd -y"
ansible cube -b -m shell -a "apt autoremove -y"

Update the OS

Simple as stated, we are going to update OS packages to the latest ones:

ansible cube -m apt -a "upgrade=yes update_cache=yes" --become

Edit /boot/firmware/cmdline.txt

Required stuff for running containers.

ansible cube -b -m shell -a "sed -i '$ s/$/ cgroup_enable=cpuset cgroup_enable=memory cgroup_memory=1 swapaccount=1/' /boot/firmware/cmdline.txt"

Iptables to see bridged traffic

Create file k3s.conf in your home directory

# add following to ~/k8s.conf
net.bridge.bridge-nf-call-ip6tables = 1
net.bridge.bridge-nf-call-iptables = 1
# We are going to use Ansible again to copy it to place, in this case /etc/sysctl.d/k8s.conf so this will load to kernel on reboot
ansible cube -b -m copy -a "src=~/k3s.conf dest=/etc/sysctl.d/k3s.conf owner=root group=root mode=0644"

Disable green led

Since we do not have an SD card in our Raspberry Pi, the green LED is going nuts. We can disable it (turn it off) with a simple command. If anybody has a way to show disk activity with the green LED related to USB disk, please let me know!

# Disable the green led ( chaning led0 to led1 you can disable power led 🙂 ) also enabling it back on, just changed echo 0 to echo 1
sudo sh -c 'echo 0 > /sys/class/leds/led0/brightness'

# Disable it for good after reboot adding paramters to /boot/firmware/config.txt
ansible cube -b -m shell -a "echo 'dtparam=act_led_trigger=none' >> /boot/firmware/config.txt && echo 'dtparam=act_led_activelow=off' >> /boot/firmware/config.txt"

My /boot/firmware/config.txt looks like this:

ubuntu@control01:~$ cat /boot/firmware/config.txt

initramfs initrd.img followkernel

# Enable the audio output, I2C and SPI interfaces on the GPIO header

# Enable the serial pins

# Comment out the following line if the edges of the desktop appear outside
# the edges of your display

# If you have issues with audio, you may try uncommenting the following line
# which forces the HDMI output into HDMI mode instead of DVI (which doesn't
# support audio output)

# If you have a CM4, uncomment the following line to enable the USB2 outputs
# on the IO board (assuming your CM4 is plugged into such a board)

There are other options that green light can do, just none that I found for USB disk :-/

# Options for dtparam=act_led_trigger=
none                No trigger
kbd-scrolllock      Keyboard scroll lock
kbd-numlock         Keyboard num lock
kbd-capslock        Keyboard caps lock
kbd-kanalock        Keyboard kana lock
kbd-shiftlock       Keyboard shift
kbd-altgrlock       Keyboard altgr
kbd-ctrllock        Keyboard ctrl
kbd-altlock         Keyboard alt
kbd-shiftllock      Keyboard left shift
kbd-shiftrlock      Keyboard right shift
kbd-ctrlllock       Keyboard left ctrl
kbd-ctrlrlock       Keyboard right ctrl
timer               Flash at 1 second intervals
oneshot             Flash only once
heartbeat           Flash like a heartbeat (1-0-1-00000)
backlight           Always on
gpio                Flash when a certain GPIO is high???
cpu0                Flash on cpu0 usage
cpu1                Flash on cpu1 usage
cpu2                Flash on cpu2 usage
cpu3                Flash on cpu3 usage
default-on          Always on
[input]             Default state
panic               Flash on kernel panic
mmc0                Flash on mmc0 (primary SD Card interface) activity
mmc1                Flash on mmc1 (secondary SD Card interface) activity
rfkill0             Flash on wifi activity
rfkill1             Flash on bluetooth activity

Remove IPv6

I don't like it, so it’s gone.

# Edit /etc/sysctl.d/99-sysctl.conf and add to the bottom:
net.ipv6.conf.all.disable_ipv6 = 1
net.ipv6.conf.default.disable_ipv6 = 1
net.ipv6.conf.lo.disable_ipv6 = 1
net.ipv6.conf.eth0.disable_ipv6 = 1

# If you do not have that file, create it and coppy contet of /etc/sysctl.conf into it first


Do a reboot of all nodes. We can speed it up a little again with Ansible, but I will target just workers. If I did a reboot on all, I might kill my main node faster than it can be executed on others… So, Ansible reboot for workers, and manually log in and reboot for control01-3.

ansible cube -b -m reboot

I hope you liked this part and got something useful. Take a break, grab some beverage and maybe for me too.

Last update: April 19, 2021