Kubernetes on Novena

Kubernetes is a high-availability clustering package. It is available for ARM, but mostly is used on x86-64. These steps detail how to use it on Novena.

Installing Kubernetes

Some steps I've taken to install Kubernetes on Debian. Adapted from https://blog.hypriot.com/post/setup-kubernetes-raspberry-pi-cluster/ for Novena.

0. Remove software that comes installed by default on Novena, but is unnecessary for servers.

apt-get remove -y --auto-remove --purge hexchat pidgin bluez bluez-tools bluez-hcidump \
       enigmail hexchat icedove iceweasel pavucontrol x11-xserver-utils xscreensaver \
       libbluetooth3 keychain xserver-xorg-video-modesetting arandr android-tools-adb \
       android-tools-fastboot android-tools-fsutils xfce4-goodies xfce4-power-manager \
       xfce4-mixer xfce4-terminal mousepad orage dbus-x11 irssi synaptic qalc libqt5gui5 \
       libqt5core5a libqt5widgets5 x11-apps x11-session-utils xbitmaps xfce4 xfce4-appfinder \
       xfce4-notifyd xfce4-session xfce4-settings xfdesktop4 xfdesktop4-data xfonts-100dpi \
       xfonts-75dpi xfonts-scalable xfwm4 xfwm4-themes xinit xorg xorg-docs-core \
       libdrm-armada2-dbg libetnaviv-dev libetnaviv-dbg novena-disable-ssp novena-eeprom-gui \
       novena-firstrun xorg-novena xserver-xorg-video-armada xserver-xorg-video-armada-dbg \
       xserver-xorg-video-armada-etnaviv

1. Update the key.

wget http://repo.novena.io/debian/pool/main/k/kosagi-repo/kosagi-repo_1.2-r1_all.deb
sudo dpkg -i kosagi-repo_1.2-r1_all.deb
apt-get update
DEBIAN_FRONTEND=noninteractive apt-get dist-upgrade
apt-get install apt-transport-https # Used for accessing the newer repos

2. At this point, it's probably a good idea to move to a SATA disk.

dd if=/dev/mmcblk0 of=/dev/sda bs=1M count=1
fdisk /dev/sda
# Re-create partitions 2 and 3.  Make partition 2 at least 6GB
# for swap (type 82), and make partition 3 the rest of the disk.
# Type "x" for Expert mode, then "i", and change the ID to 0x4e6f7653.
# Then "r" to return to the main menu, and "w" to write it to disk.
mkfs.ext4 /dev/sda3 # or install btrfs-tools or xfsprogs and make a different type
mount /dev/sda3 /mnt
rsync -avxHAX --progress / /mnt/
novena-eeprom -w -f es8328,pcie,gbit,hdmi,eepromoops,sataroot
reboot

3. Install Docker (borrowed from https://github.com/hypriot/image-builder-odroid-c1/blob/master/builder/chroot-script.sh). If you need a specific version, you can list them with:

apt-cache madison docker-engine

wget -q https://packagecloud.io/gpg.key -O - | apt-key add -
echo 'deb https://packagecloud.io/Hypriot/Schatzkiste/debian/ jessie main' > /etc/apt/sources.list.d/hypriot.list
apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 2C52609D
echo 'deb [arch=armhf] https://apt.dockerproject.org/repo debian-jessie main' > /etc/apt/sources.list.d/docker.list
apt-get update
DEBIAN_FRONTEND=noninteractive apt-get install \
   docker-engine=1.13.1-0~debian-jessie \
   docker-compose="1.9.0-23" \
   docker-machine="0.9.0-39"

4. Install kubeadm

curl -s https://packages.cloud.google.com/apt/doc/apt-key.gpg | apt-key add -
echo "deb http://apt.kubernetes.io/ kubernetes-jessie main" > /etc/apt/sources.list.d/kubernetes.list
apt-get update && apt-get install -y kubeadm

5. Run kubeadm on the master device to start the cluster. As of this moment, we need to skip preflight checks due to changes in Docker version numbering.

kubeadm init --pod-network-cidr 10.244.0.0/16 --skip-preflight-checks

It will output something like the following:

Your Kubernetes master has initialized successfully!

You should now deploy a pod network to the cluster.
Run "kubectl apply -f [podnetwork].yaml" with one of the options listed at:
    http://kubernetes.io/docs/admin/addons/

You can now join any number of machines by running the following on each node:

kubeadm join --token=964a50.d8053ed3de195a11 10.0.245.169

6. Join the cluster from other machines. We still need to skip preflight checks.

kubeadm join --token=964a50.d8053ed3de195a11 --skip-preflight-checks 10.0.245.169

7. Ensure cAdvisor doesn't start up. It's nice to have, but it leaks lots of information.

printf '[Service]\nEnvironment="KUBELET_EXTRA_ARGS=--cadvisor-port=0"\n' > /etc/systemd/system/kubelet.service.d/05-disable-cadvisor.conf
systemctl daemon-reload
systemctl restart kubelet

8. Install Flannel on the Master, which will manage the network for us.

curl -sSL https://rawgit.com/coreos/flannel/master/Documentation/kube-flannel.yml | sed "s/amd64/arm/g" | kubectl create -f -

9. Wait for all pods to be Running:

kubectl get po --all-namespaces

10. As of this post, networking is broken. To fix it, break the firewall for the Docker interface:

 sudo iptables -A FORWARD -i cni0 -j ACCEPT
 sudo iptables -A FORWARD -o cni0 -j ACCEPT